THEORY OF EQUATIONS.

DUBLIN UNIVERSITY PRESS SERIES.

THE

THEORY OF EQUATIONS
WITH AN

INTRODUCTION TO THE THEORY OF BINARY

ALGEBRAIC FORMS.
BY

WILLIAM SNOW BURNSIDE,

FELLOW OF TRINITY COLLEGE, DUBLIN

M.A.,
;

ERASMUS SMITH'S PROFESSOR OF MATHEMATICS THE UNIVERSITY OF DUBLIN:

AND

IN

ARTHUR WILLIAM PANTON, M. A, FELLOW AND TUTOR, TRINITY COLLEGE, DUBLIN;

DONEGAL LECTURER
IN

MATHEMATICS.

SECOND EDITION.
DUBLIN: HODGES, FIGGIS, &
CO.,

GRAFTON-STREET.

LONDON LONGMANS, GREEN, &

:

CO.,

PATERNOSTER-ROW.

1886.

DL'HLIN

PRINTED AT THE UNIVERSITY PHKSS,

BY PONSONUY AND WELDRICK.

PREFACE TO TIE FIRST EDITION.

WE have
the

endeavoured in the present work to combine some of

of

modern developments
ten Chapters contain

Higher Algebra with the subjects

usually included in works on the Theory of Equations.

first

The

all

the propositions ordinarily found

in elementary treatises on the subject.

In these Chapters we

have not hesitated to employ the more modern notation wherever

it

appeared that greater simplicity or comprehensiveness

could be thereby obtained.

Regarding the algebraical and the numerical solution of

equations as essentially distinct problems,

we have purposely
and biquadratic

omitted in Chap. VI. numerical examples in illustration of the

modes

of solution there given of the cubic

equations.
of

Such examples do not render

;

clearer the conception

an algebraical solution

and, for practical purposes, the

algebraical formula

may

be regarded as almost useless in the

case of equations of a degree higher than the second.

In the treatment
as well as in the
tions, a

of Elimination

and Linear Transformation,

of

more advanced treatment

is

Symmetric Func"We have

knowledge of Determinants

indispensable.

vi

Preface
it

to the

First Edition.

found

necessary, therefore, to give a Chapter

on

this subject.
intelli-

It has been our

aim

to

make

this
;

Chapter as simple and

gible as possible to the beginner

and

at the

same time

to omit

no proposition which might be found useful in the application

of this calculus.

For many

of the

examples in this Chapter, as

well as in other parts of the work,

we

are indebted to the kind-

ness of Mr. Cathcart, Fellow of Trinity College.

We

have approached the consideration of Covariants and

Invariants through the

medium

of the functions of the diffethis

rences of the roots of equations

plest

appearing to us the sim-

mode

of presenting the subject to beginners.

We

have

attempted at the same time to show

how

this

mode

of treatment

may
on

be brought into harmony with the more general problem of

the linear transformation of algebraic forms.

this subject

In the Chapters

we have

confined our attention to the quadratic,

cubic,

and quartic; regarding any complete discussion of the covariants and invariants of higher binary forms as too diffi-

cult for a

work

like the present.

Of

the works which have afforded us assistance in the

more

elementary part of the subject,

the Traitt d* Alyebre of
late Professor

we wish

to

mention particularly
the writings of the

M. Bertrand, and

Young*

of Belfast,

which have contributed so

much

to

extend and simplify the analysis and solution of

numerical equations.

In the more advanced portions of the subject we are

in-

debted mainly, among published works, to the Lessons Introto the

Minimi Hiyhcr Algebra of Dr. Salmon, and the

Tli<T>/
1. 1

atiil

X<, Int
'

ion

of Algebraical

7v//w^'o//.v,

Jiit/mHlratic
r'.'jiHtt'HHi*

Equations,

London,

London, 1835; Analysis and 1842; and Theory and

of

tin-

II iff /HI-

Orders, London, 1843.

Preface
Theorie
in

to the

First Edition.

vii

der

bindren
to

algebraischen

Formen
des

of Clebsch
binaires

and

some degree

the

Theorie

Formes

of the

Chev. F.

Faa De Bruno.

We must record also our

obligations

in this department of the subject to

Mr. Michael Eoberts, from

periodicals,

whose Papers in the Quarterly Journal and other

and from whose

professorial

lectures

in the University of

Dublin, very great assistance has been derived.

Many

of the

examples

also are

taken from Papers

set

by him

at the

Uni-

versity Examinations.

In the Chapter on the Complex Variable we have followed

closely the treatment of

imaginary quantities given by M.

Briot in his Lemons d'Algebre.

In connexion with various parts of the subject several other works have been consulted, among which may be
mentioned the
treatises

on Algebra by

Serret,

Meyer Hirsch,

and Rubini, and papers in the mathematical journals by Boole, Cay ley, Sylvester, Hermite, and others.

We

have, in the last place, to express our thanks to Mr.

Eobert Graham, of Trinity College, Dublin, who has read the

proof sheets, and verified most of the examples.

His thorough acquaintance with the subject has been invaluable to us, and many improvements throughout the work are owing to suggestions

made by him.

TRINITY COLLEGE,
September, 1881.

PEEFACE TO THE SECOND EDITION.

THE

chief alterations in the present edition will be

found in

the Chapter on Determinants, which has been considerably en-

and in Chap. XVI., on Transformations, to which two new sections have been added. The former of these contains a
larged
;

dkcussion of Hermite's theorem relating to the limits of the

roots of

an equation

and in the

latter

we have given an
which the Theory

account of the method of transformation from a system of two

to a system of three variables,

by means

of

of

Covariants and Invariants of Binary

Forms

receives

simple geometrical illustration.

TRINITY COLLEGE,
December, 1885.

NOTE.

The

first

ten Chapters of this work

may

be regarded as forming an
first

elementary course.

In reading these Chapters for the

time,

Students are

recommended

to

omit Art. 53 of Chap. V., and to confine their attention in

Chap. VI. to Arts. 55, 50, 57, 61, 62, and 63.

TABLE OF CONTENTS,

INTRODUCTION.
Art.
1.

Page
Definitions,
1

2.
3.

Numerical and algebraical equations,

Polynomials,

2
3

CHAPTER

I.

GENERAL PROPERTIES OF POLYNOMIALS.

4.

Theorem

relating to polynomials

when

the variable receives large values,

5.
6.

Similar theorem

when

the variable receives small values,

to

...
by a

5
6

Change of form of a polynomial corresponding

tion of the variable.

an increase or diminu8 9
is

Derived functions,
divided

7.
S.

Continuity of a rational integral function, Form of the quotient and remainder when a polynomial
binomial,

10
.

9.

Tabulation of functions,

12

10.
11.

Graphic representation of a polynomial, Maxima and minima values of polynomials,

.13
17

CHAPTER

II.

GENERAL PROPERTIES OF EQUATIONS.

12, 13, 14.

Theorems relating

to the real roots of equations,

15. Existence of a root in the general equation. 16. Theorem determining the number of roots of
17. 18.

Imaginary roots, an equation,

.19 .21 .22

25

Equal

roots,
pairs,

Imaginary roots enter equations in

Ij^Descartes' rule of signs for positive roots,

......

26

28

Table of Contents.
Art.

Descartes' rule of signs for negative roots, of imaginary roots, <|^ Use of Descartes' rule in proving the existence
22.

......
.
.

Page 30
30

Theorem
Examples,

relating to the substitution of

variable,

............ ... .........

CHAPTER
III.

two given numbers for the

31

32

RELATIONS BETWEEN THE BOOTS AND COEFFICIENTS OF EQUATIONS, WITH APPLICATIONS TO SYMMETEIC FUNCTIONS OF THE EOOTS.
23. Relations

between the roots and

coefficients.

Theorem,

.35
36

24. Applications of the theorem, 25. Depression of an equation

26. 27.

when a
roots,

relation exists

between two of

its roots,

42
46
48 53
54

The cube

roots of unity,

.43

Symmetric functions of the

Examples, 28. Theorems relating to symmetric functions,

Examples,

CHAPTER

IV.

TEANSFOEMATION OF EQUATIONS.
29. Transformation of equations, 30. Eoots with signs changed, 31. Eoots multiplied
32. Reciprocal roots

.60
60
.
. .

by a given

quantity,

.61
62

and reciprocal equations, increase or diminish the roots by a given quantity, 34. Removal of terms,
33.

To

35.

Binomial

coefficients,

.....
...

.64 ,67 .68

71
73,

36.
37. 38.

The The

cubic,

biquadratic,

39. Transformation

40.

Homographic transformation, by symmetric functions, Formation of the equation whose roots are any powers of the roots of the
proposed equation,

75 76

73

41. Transformation in general, 42. Equation of squared differences of a cubic, 43. Criterion of the nature of the roots of a cubic, 44. Equation of differences in general,

,80
81
84

...

84

.86

Table of Contents.

xi

CHAPTER
Art.

V.

SOLUTION OF RECIPROCAL AND BINOMIAL EQUATIONS.

Page
90
Propositions embracing their leading general
45. Reciprocal equations,

46-52. Binomial equations.

properties,

.92
.

53.

The

special roots of the equation

0,

54. Solution of binomial equations

by

circular functions,

....
.
.

95
98 100

Examples,

CHAPTER

VI.

ALGEBRAIC SOLUTION OF THE CUBIC AND BIQUADRATIC.

55.
56.

On

The

the algebraic solution of equations, algebraic solution of the cubic equation,

.103
106
107 109

57. Application to numerical equations, 58. Expression of the cubic as the difference of
59.

two cubes,

....
.
.

Solution of the cubic

by symmetric functions of the

between two roots of a cubic,

roots,

.111
112

Examples,
60.

Homographic

relation

.118
.

61. First solution

by

radicals of the biquadratic.

Euler's assumption,

11!

Examples, 62. Second solution by radicals of the biquadratic,

12o
. .

.125
127
131

63. Resolution of the quartic into its quadratic factors.

Ferrari's solution,

64. Resolution of the quartic into its quadratic factors.

Descartes' solution,
.
.

65. Transformation of the biquadratic into the reciprocal form, 66. Solution of the biquadratic by symmetric functions of the roots,
67.

133

Equation of squared differences of a biquadratic,

Examples,

68. Criterion of the nature of the roots of a biquadratic.

.

.... ....
.

.137
140
142 144

CHAPTER

VII.

PROPERTIES OF THE DERIVED FUNCTIONS.

69. Graphic representation of the derived function,
. .

.152
.

Theorem

relating to the

maxima and minima

.
.

values of a polynomial,

153

5J. Rolle's theorem.

Corollary,

72. Constitution of the derived functions,

73.

.155 .155
156

Theorem

relating to multiple roots,

74. Determination of multiple roots,

75. 76.

.157
159,

Theorems relating to the passage of the variable through a root of

the equation,

160
161

Examples,

xii

Table of Contents.

CHAPTER
Art.

VIII.

LIMITS OF THE BOOTS OP EQUATIONS.

7. Definition of
limits,

163
I.,

Limits of roots.
79. Limits of roots.
10.

Prop.

163
164

Prop. II.,

Practical applications,

82. Inferior limits,

U. Newton's method of finding limits. Prop. III., and limits of the negative roots

....

166

168
169

83. Limiting equation?,

170
171

Examples,

CHAPTER
General explanation,

IX.

SEPARATION OF THE KOOTS OF EQUATIONS.

84.

.172
172

85.

Theorem of Fourier and Budan,

theorem to imaginary roots, from the theorem of Fourier and Budan,

86. Application of the theorem, 87. Application of the 88. Corollaries

.175
177
. . .

.180
181

Sturm's theorem,
90. Sturm's theorem.

Equal

roots,

186

189. 91. Application of Sturm's theorem,

92. Conditions for the reality of the roots of

an equation,

....
. .

189

193
194

93. Conditions for the reality of the roots of the biquadratic,

Examples,

195

CHAPTER
94. Algebraical

X.

SOLUTION OF NUMERICAL EQUATIONS.

and numerical equations, Theorem relating to commensurable roots, Newton's method of divisors, Application of the method of divisors, Method of limiting the number of trial-divisors,

95. 96. 97. 98.

...... .....
.... ....
.
.
.

197
198

199

200

203
204
207 209 213 217
220 223
224 227

99. Determination of multiple roots,

100. Newton's
101.

102.

method of approximation, Homer's method of solving numerical equations, Principle of the trial-divisor in Homer's method,
where

103. Contraction of Horner's process, 104. Application of Horner's method to cases

10-"j.

roots are nearly equal,

Lagrange's method of approximation, 106. Numerical solution of the biquadratic by Descartes' method,
i'lri,

Table of Contents.

xiii

CHAPTER
Art.

XI.

DETERMINANTS.
107. Elementary notions and definitions,

Page 229

108. Rule with regard to signs, 109-112. Elementary propositions relating to determinants.

232
Props.
I. -IV.,

113. Minor determinants.

Definitions,

.......
.

234
239

114. Development of determinants,

115. Laplace's development of determinants, 116. Development of a determinant in products of leading constituents,
117. Expansion of a determinant
.

239

244
245

by products in

pairs of the constituents of

a row and column,

Prop. V., . 119-120. Further propositions. Prop. VI. and Prop. 121. Multiplication of determinants. Prop. VIII.,
122. Second proof of Prop. VIII.,
118. Addition of determinants.

VI I.,

....250
248
.

247

255

256 259 262

263

123. Rectangular arrays, 124. Solution of a system of linear equations, 125. Linear homogeneous equations,
126. Reciprocal determinants,
127. Symmetrical determinants,
128.

264
267

Skew-symmetric and skew determinants, 129. Theorem relating to a determinant whose leading
Miscellaneous Examples,
. .

269
first

minor vanishes,

273
276

CHAPTER

XII.

SYMMETRIC FUNCTION'S OF THE ROOTS.

130. Newton's theorem on the

sums of powers of

roots.

Prop.

I.,

'289

131. Expression of a rational symmetric function of the roots in terms of the

coefficients.

Prop. II.,

291

132. Further proposition relating to the expression of in terms of the coefficients. Prop. III., 133. Expression of the coefficients in terms of

sums of powers

of roots

293
of roots,
.

sums of powers

294

134. Definitions of order and weight of symmetric functions, and theorem relating to the former,
135. Calculation of symmetric functions of the roots, 136. Brioschi's differential equation connecting the sums of the powers of the roots and the coefficients, 137. Derivation of

304

new symmetric

functions from a given one,

306
307

138. Equation of operation, 139. Operation involving the

sums of the powers of the

roots.

Theorem,

309
310

Miscellaneous Examples,

xiv

Table of Contents.

CHAPTER
Art.

XIII.

ELIMINATION.
140. Definitions,
141. Elimination

Pa &e 318

by symmetric

functions,

319

142. Properties of the resultant, 143. Euler's method of elimination, 144. Sylvester's dialytic method of elimination, 145. Bezout's method of elimination, 146.

320
322 323
324

The common method

of elimination,

329

147. Discriminants, 148. Determination of a root

common

to

149. Symmetric functions of the roots of

two equations, two equations,

.... ....

331

333 334

Miscellaneous Examples,

33&

CHAPTER XIV.
COVARIANTS AND INVARIANTS.
150. Definitions,

...

338
339
341

151. Formation of covariants

and invariants, 152. Properties of covariants and invariants,

Examples, 153. Formation of covariants by the operator D., 154. Roberts' theorem relating to covariants,
155.

343

344 346
.
.

Homographic transformation applied

tion,

to the theory of covariants,

347

156. Reduction of homographic transformation to a double linear transforma-

349

157. Properties of covariants derived

158-161. Propositions relating to of quantics transformed by a linear transformation, Miscellaneous Examples,

from linear transformation, . .351 the formation of invariants and covariants

. .

.354
360

CHAPTER XV.
COVARIANTS AND INVARIANTS OF THE QUADRATIC, CUBIC, AND QUARTIC.
162. 163.

The quadratic, The cubic and its

36&
covariants,
.

164.
165.

Number
The

of covariants
;

quartic

its

and invariants of the cubic, covariants and invariants,

....

366

369

.371
372

166. Quadratic factors of the sextic covariant,

Table of Contents.
Art.

xv
Page

167. Expression of the Hessian

variant,

by the quadratic

factors of the sextic co-

374

168. Expression of the quartic itself

by the quadratic

factors of the sextic

covariant,
169. Resolution of the quartic,

374

376

170.
171.

The

invariants and covariants of

of covariants

KU-\HX

Number

and invariants of the quartic,

....

377 379

Miscellaneous Examples,

382

CHAPTER XYL

TRANSFORMATIONS.
SECTION
172. Theorem,
I.

TSCHIRNHATJSEN'S TRANSFORMATION.
,

385

173. Formation of the transformed equation,

174. Second

method of forming the transformed equation,

175. Tschirnhausen's transformation applied to the cubic, 176. Tschirnhausen's transformation applied to the quartic. 177. Reduction of the cubic to a binomial form

.... ....
Theorem,
.

387

388 389
390

by Tschirnhausen's

transfor-

mation,
178. Reduction of the quartic to a trinomial form

391

by Tschirnhausen's

trans-

formation,
179.

392
third,

Removal of the second, n th degree,

theorem,

and fourth terms from an equation of the

392

180. Reduction of the quintic to the

sum

of three fifth powers.

Sylvester's

395 396

181. Quartic s transformable into each other,

Miscellaneous Examples,

398

SECTION II.
182.

HERMITE'S THEOREM.
sum
of squares,
.

Homogeneous function
.

of second degree expressed as

401

183. Hermite's theorem,

403 406
409

184. Sylvester's forms of Sturm's functions,

Examples,

SECTION III.

GEOMETRICAL TRANSFORMATIONS.

185. Transformation of binary to ternary forms,

186.

The

quartic and

its

covariants treated geometrically,

....

411

414

187. Determination of the proper ternary form,

416
418

Examples,

xvi

Table of Contents.

CHAPTER

XVII.

THE COMPLEX VARIABLE.

Art.

188. Graphic representation of imaginary quantities, 189. Addition and subtraction of imaginary quantities,
190. Multiplication 191.

and division of imaginary

quantities,

.... ....
.

Page 422

423
424 425

The complex

variable,

. . . 192. Continuity of a function of the complex variable, 193. Variation of the argument of the function corresponding to the descrip-

.427

tion of a small closed curve

194. Cauchy's theorem relating to the

. .428 by the complex variable, number of roots comprised within a

.

plane area,
195. Proof of the fundamental theorem that every equation of the

430

n th degree
431

has n roots real or imaginary,

NOTES.
A. Algebraic solution of equations, B. Solution of numerical equations,
C. Determinants,
.

.433
437
441

D. The propositk i that every equation has a

root,

442

THEORY OF EQUATIONS.

INTRODUCTION.
1.

Definitions.
is

Any

mathematical expression involving a

quantity

called & function of that quantity.

employed mainly with such algebraical functions as are rational and integral. By a rational function of a quantity is meant one which contains that quantity in a rational form only that is, a form free from fractional indices or radical By an integral function of a quantity is meant one in signs. which the quantity enters in an integral form only that is,
;
;

"We

shall be

never in the denominator of a fraction.

sion, for

The following
is

a positive integer, and integral algebraical function of x

is
:

example, in which n

expresa rational

~* ax? + bzn

cxn

~*

+ ...;.,+ kx + L

It is to be observed that this definition has reference to the

variable quantity x only, of which the expression is a function. The several coefficients , &, c, &c., may be irrational or fractional,

and the function

or

still

remain rational and integral in

x.

A function of x is represented for brevity by F(x},f(x),

some such symbol.

$(#),

The name polynomial

to express the fact that it

is

given to the algebraical function is constituted of a number of terms

containing different powers of x connected by the signs plus or

:_>

Introduction.

minus.

For

nomial

may

certain values of the variable quantity x one polyto another differently constituted. become

equal

The
tion
;

of such a relation is called an equaalgebraical expression and any value of x which satisfies this equation is called a

root of the equation.

The determination
by bringing
:

of all possible roots

constitutes the complete solution of the equation.

It is obvious that,

all

the terms to one side,

we
x

arrange any in the following manner

a %n +
i

may

equation according to descending

powers of

a?""

+ a 2 xn

~z

+....+ cin-ix +

a n = 0.

highest power of x in this equation being n, it is said to th be an equation of the n degree in x. For such an equation we The suffix shall, in general, employ the form here written.

The

ficient

attached to the letter a indicates the power of x which each coefaccompanies, the sum of the exponent of x and the suffix

of a being equal to
if all its

we

n for each term. An equation is not altered terms be divided by any quantity. may thus, if n make the coefficient of x in the above , please, dividing by

We

equation equal to unity.

It will often be found convenient to

make

this supposition

and in such

cases the equation will be

written in the form

Xn + p^Xn~^ + p z X
J

tl ~*

+ p n-l%+Pn =

0.

equation is said to be complete involving x in all its powers from n to

An

when
0,

it

contains terms

and

incomplete

some

of the terms are absent

or, in

other words,

the coefficients

l9

p 2)

&e.,
ar,

which does not contain

tion
is

is

are equal to zero. called the absolute term.

when when some of The term p n

,

An

equa-

numerical, or algebraical, according as its coefficients are numbers, or algebraical symbols.

2.

Numerical and Algebraical Equations.

and physical

In many

researches in both mathematical

science the final

mathematical problem presents

itself in the form of an equation on whose solution that of the problem depends. It is natural, therefore, that the attention of mathematicians should have been

Numerical and Algebraical Equations.

at

an early stage in the history of the science directed towards

The science of the Theory of Equainquiries of this nature. has tions, as it now stands, grown out of the successive attempts
of mathematicians to discover general

methods for the solution

of equations of any degree. When the coefficients of an equation are given numbers, the problem is to determine a numerical
value, or perhaps several different numerical values, which will In this branch of the science very great satisfy the equation. has been made and the best methods hitherto advanced ; progress
for the discovery, either exactly or approximately, of the
rical values of the roots will

nume-

be explained in their proper places

in this work.

Equal progress has not been made in the general solution of equations whose coefficients are algebraical symbols. The student is aware that the root of an equation of the second degree,
whose
coefficients are

such symbols,

may

be expressed in terms

of these coefficients in a general formula ; and that the numerical roots of anyjparticular numerical equation may be obtained

substituting in this formula the particular numbers for the symbols. It was natural to inquire whether it was possible to

by

discover

any such formula for the solution of equations of higher Such results have been attained in the case of equadegrees. It will be shown that tjons of the third and fourth degrees.
in certain cases these formulas fail to supply the solution of a numerical equation by substitution of the numerical coef-

general symbols, and are, therefore, in this respect inferior to the corresponding algebraical solution of the quadratic.
ficients

for the

Many attempts have been made to arrive at similar general formulas for equations of the fifth and higher degrees; but it may now be regarded as established by the researches of modern
analysts that it is not possible by means of radical signs, and other signs of operation employed in common algebra, to express the root of an equation of the fifth or in terms of the coefficients.
3.

any higher degree

it is

Polynomials.
<T

From

the preceding observations

B2

Introduction.

plain that one important object of the science of the Theory of Equations is the discovery of those values of the variable quantity
zero.

x which give to the polynomial f(x) the particular value In attempting to discover such values of x we shall be led

into

many

inquiries concerning the values

assumed by the polyshall, in fact, see

nomial for other values of the variable.

We

in the next Chapter that, corresponding to a continuous series of values of x varying from an infinitely great negative quantity (- oo
)

to

an

infinitely great positive quantity (+00), f(x]

will

assume

such variations
nomials.
fact,

The study of a very important part of the theory of polyThe general solution of numerical equations is, in
also values continuously varying.
is
;

and by examining the values assumed by the polynomial for certain arbitrarily assumed values of the variable, we shall be led, if not to the root itself, at least to an indication of the neighbourhood in which it exists, and within which our further approximation must be carried on.
a tentative process

A polynomial
to

is

sometimes called a quantic.

It is convenient

have

distinct

degrees.

names The terms quadratic

for the quantics of various successive

(or quadric), cubic, biquadratic (or

quartic), quintic, sextic, &c., are

used to represent quantics of the

2nd, 3rd, 4th, 5th, 6th, &c., degrees ; and the equations obtained by equating these quantics to zero are called quadratic, cubic,
biquadratic, &c., equations, respectively.

CHAPTER

I.

GENERAL PROPERTIES OF POLYNOMIALS.

4.

IN tracing the changes

of value of a polynomial correspond-

ing to changes in the variable, we shall first inquire what terms in the polynomial are most important when values very great

This inquiry will form the or very small are assigned to x. the and of succeeding Articles. present subject the polynomial in the form Writing
(

a^

az l

_!

an 1

it is

plain that
oo
.

its

towards

The

value tends to become equal to a^x" as x tends following theorem will determine a quantity
effect of
this, or of any greater quantity, n making the term a Q x exceed the In what follows we suppose a to be

such that the substitution of

for
a?

will

have the

sum

of all the others.

positive;

and in general in the treatment

is

of polynomials

and

equations the highest term

positive sign.

supposed to be written with the

Theorem.
a n xn
the value

If in

the polynomial
1

a^'

+ a 2 af*~z +

+ a n _iX + a n

+
a

1, or

any greater

value, be substituted for x, where au

is

that one of the coefficients

a^ # 2

>

whose numerical value

is

yreatest, irrespective of of x will exceed the sum of all

sign, the

term containing the highest power the terms which follow.

The

inequality
an

6
is satisfied

General Properties of Polynomials.

by any value
a^x
n

of x
n~ l

which makes
+ %n
~2

>

ak (x

+x+

1),
1? z,
. . .

where a k

is

the greatest

without regard to sign. the brackets, we have

/yttl

ffw-i, among the coefficients a a series within the Summing geometric

,

a*x

>ak -----,

"I

x-1

or xn

>

sj.

(x

1),

(#-1)

which
that
is

is satisfied if

a Q (x - 1) be > or = a^

x > or =
Co
is

1.

The theorem here proved

coefficients of the

polynomial are

when the given numbers, a number such

useful in supplying,

that

when x

receives values nearer to

oo

the polynomial will

preserve constantly a positive sign. If we change the sign of x y the first term will retain its sign if n be even, and will become

negative if n be odd so that the theorem also supplies a negative value of a, such that for any value nearer to - oo the
;

polynomial will retain constantly a positive sign if n be even, and a negative sign if n be odd. The [constitution of the polyis, in general, such that limits much nearer to zero than those here arrived at can be found beyond which the function

nomial

preserves the same sign ; for in the above proof we have taken the most unfavourable case, viz. that in which all the coefficients

except the first are negative, and each equal to ak whereas in general the coefficients may be positive, negative, or zero.
;

Several theorems, having for their object the discovery of such closer limits, will be given in a subsequent Chapter. 5. now proceed to inquire what is the most important

We
;

term in a polynomial when the value] of x is indefinitely diminished and to determine a quantity J such that the substitution
of this, or of

any smaller quantity,

If in
the

for x will 'have the effect of

giving such term the preponderance.

Theorem.

polynom i<d

Theorem.
the value
is

or any smaller value, be substituted for x, where a k

the greatest coefficient exclusive of a n , the term a n will be

nume-

rically greater than the

sum of all
x = J
;

the others.

To

prove

this, let

then by the theorem of Art.

the coefficients #, a ly
.

4,

a k being

now

the greatest

among

a n . ly
of

without regard to sign, the value

(in

1,

or

any greater value

y, will

make
an yn

> a n ^y n ~ + a n .2 y n ^ +
l

that

is,

an >

ff w _x

11
+

+ a,y + a Qy
1

a n _,

yr
less

hence the value

or

any

value of

x, will

make

This proposition
follows
:

is

often stated in a different manner, as

Values so small

may

be assigned to

x as

to

make

th<

polynomial

less

than any assigned quantity. This statement of the theorem follows at once from the above

proof, since a n

may

There
follows
:

is also

When

be taken to be the assigned quantity. another useful statement of the theorem, as the variable x receives a very small value the
,

of the polynomial

is

the

same as

the sign of its first term

This appears by writing the expression in the form

.
.

+a x

tl

~1
} ;

for

when

a value sufficiently small

is

given to

x,

the numerical

value of the term a n _i exceeds the

sum

of the other terms of

will consequently

expression within the brackets, and the sign of that expression depend on the sign of # w _i.

General Properties of Polynomials.

to

6. Change of Form of a Polynomial corresponding an increase or diminution of the Variable. Derived Functions. We shall now examine the form assumed by the

polynomial when x + h is substituted for x. If, in what follows, h be supposed essentially positive, the resulting form will correspond to an increase of the variable and the form corresponding
;

to a diminution of x will be obtained

from

this

by changing the
A), or

sign of h in the result.

When
(#

is

changed
*

to
1

x + h,f(x) becomes /(# + + a ~ (X +

2

+ h)* + a ( X +

^)"~

^)"~

#?*-!

(#

h)

+ Un-

Let each term of this expression be expanded by the binomial theorem, and the result arranged according to ascending powers then have of h.

We
+

a xn

#w

~i

tf 3

#w

~2

+
aJ

fl w _ 2

#2

4-

a n-\x + a n
~z

+ h {na xn

~l

(n

w -3

1)

+ (n - 2) a 2 x n

n i
.

It will be observed that the part of this expression independent of h isf(x) (a result obvious d priori) and that the successive coefficients of the different powers of h are functions of x of
9

degrees diminishing by unity. It will be further observed that the coefficient of h may be derived from/(#) in the following

manner:

of x in that term,

Let each term in/(#) be multiplied by the exponent and let the exponent of x in the term be
;

diminished by unity, the sign being retained the sum of all the terms of f(x) treated in this way will constitute a polynomial of dimensions one degree lower than those of f(x). This polynomial is called the first derived function of /(a?). It is usual to
represent this function

by the notation

/'(#).

The

coefficient

Derived
A2
of
^

Fund ions.

may

be derived from/'(.r) by a process the same as that

employed in deriving /'(#) from/(#), or by the operation twice performed on/(#). This coefficient is represented by /"(#), and In like manner the is called the second derived function of /(#).
succeeding coefficients

may

all

be derived by successive opera-

tions of this character; so that,

employing the notation here

:

indicated,

we may
A)

write the result as follows

f(x +
It

=/(*) +/'(*) h

+'Y%

If

*+...+

flo h".

may

does not alter /(a? + A), the expansion

be observed that, since the interchange of x and h may also be written in the

form
'/(x + A) -/(A) +/'(/<)*

We

shall in general

employ the notation here explained;

but on certain occasions when

it is necessary to deal with derived functions beyond the first two or three, it will be found more convenient to use suffixes instead of the accents here employed.

The expansion
h)

will then be written as follows

EXAMPLE.
Find the result of substituting x Here
/(*)

h for x in the polynomial 4z 3 '+ Qx 2

-7x +

4.

=4^ + 6^-7^ + 4,
12s- 7,

f'(x) =12o;2+

f"(x) =24^;+ 12,

f'"(,v}

= 2;
i

and the

result is

The student may

7.

verify this result

by

direct substitution.

If in a rational

Continuity of a Rational Integral Function of a\ and integral function f(x) the value of x be

10

General Properties of Polynomials.

to vary, by indefinitely small increments, from one quana to a greater quantity 6, we proceed to prove that f(x) at tity the same time varies also by indefinitely small increments ; in

made

other words, that/(#) varies continuously with x. Let x be increased from a to a + h. The corresponding incre-

ment

of f(x) is

/( + *)-/();

and

this is equal,

by Art.

6, to

in which expression all the coefficients f(a), f"(a), &c., are finite Now, by the theorem of Art. 5, this latter expresquantities.

may, by taking h small enough, be made to assume a value than any assigned quantity so that the difference between f(a + h} and/(#) may be made as small as we please, and will ultimately vanish with h. The same is true during all stages of
sion
less
;

the variation of x from a to b

thus the continuity of the funcit

tion/^)

is

established.
is

It is to be observed that
increases continuously

not here proved that/ (x)

.

or diminish, or at

from f(d) to f(b) It may either increase one time increase, and at another diminish
;

but the above proof shows that it cannot pass per saltmn from one value to another and that, consequently, amongst the
;

values assumed
b

by /(^) while x
i

must be included

all

increases continuously from a to The sign values between /(a) and/(6).

for it appears

of f'(d) will determine whether f(x) is increasing or diminishing; by Art. 5 that when h is small enough the sign of

the total increment will depend on that of f'(a] h. observe that when (a) is positive f(x) is increasing with x

We
;

thus

and

when

(a) is negative

f(x)

is

diminishing as

increases.

of the Quotient and Remainder divided by a Binomial. Let the Polynomial when
8.

Form

when a
quotient,

is

Form of
is

Quotient and Remain

divided

by x -

h,

be

This

we

shall represent

by

Q,

and the remainder by E.

:

We

have then the following equation

The meaning of this equation is, that when Q is multiplied - h, and x added, the result must be identical, term for term, by with /(a?). In order to distinguish equations of the kind here

explained from equations which are not identities, it will often be found convenient to use the symbol here employed in place
of .the usual

symbol of equality.

The right-hand

side of the

identity

is

- hbj
Equating the
coefficients of

-lib n ^\

- hb n .i.

x on both sides, lowing series of equations to determine b 09 b l9 b*

9

we
. .

get the folb n . ly

b
bi

= a09 =
b

h +

0i

b n -i= bn_Ji

b&.\h

+ an

These equations supply a ready method of calculating in

succession the coefficients b 09 b l9 &G. of the quotient, and the remainder R. For this purpose we write the series of operations

in the following

manner

bji,
bi 9

62,

63 ,

....

b n -i 9

R-

In the

first line

are written

down

the successive coefficients

12
of f(x).

General Properties of Polynomials.

The

first

term in the second line

is

is

obtained

plying a (or b 09 which

equal to

it)

by
it

h.

by multiThe product b h is
is

placed under i, and then added to term bi in the third line. This term,
plied in
its

in order to obtain the

when

obtained,
.

multi-

turn by A, and placed under a 2 The product is added to #2 to obtain the second figure b z in the third line. The

repetition of this process furnishes in succession all the coefficients of the quotient, the last figure thus obtained being the

remainder.

A.

few examples will make

EXAMPLES.

this plain.

1.

Find the quotient and remainder when 3* 4 - 5# 3 + 10*2 + llx

is

- 61

is

divided

by x 3. The calculation

arranged as follows

3-5
9

10 12

11

61.

66
77

231.
170.

22

Thus the quotient is 3* 3 + 4*2 + 22* + 77, and the remainder 170. 2 3 2. Find the quotient and remainder when * + 5* + 3* + 2 is divided hy x Ana. Q
3.

1.

=
by x - 5. must be taken
to supply

Find Q and

R when x 5 - 4** + 7*3 - llx - 13

term in a polynomial
is

is

divided

N.B.

When any

absent, care

the place of

its coefficient

by zero in writing down the

first line

coefficients of /(*).

In

this

example, therefore, the series in the

;i

will be

-4

-11

-is.
;

Ana. Q = ** + x*
4.

Find Q and It when *9 + 3* 7 - 15*2 + 2 is Ana. C} = * 8 + 2* 7 + 7* 6 +14* 5 + 28* 4 + 56*3 + 112*2 + 209* + 418;

+ 12** + 60* + 289 divided by x - 2.

=1432.

.#

= 838.

5.

Find Q and

It

when x 5 + *2 - 10* + 113 is divided by * + 4. Ans. Q = x*- 4* 3 + 16*2 -63* + 242;

= -855.

9. Tabulation of Functions. The operation explained in the preceding Article affords a convenient practical method of calculating the numerical value of a polynomial whose coef-

ficients are

given numbers

when any number is

substituted for x.

For, the equation

f(x)^(x-h)Q + K,
since its

two members are identically equal, must be satisfied when any quantity whatever is substituted for x. Let x = /?,

Tabulation of Functions.

13

x h being = 0, and Q remaining finite. Hence 9 then/(/*) the result of substituting h for x in/(#) is the remainder when - h, and can be calculated /'(#) is divided by x rapidly by the process of the last Article.

For example, the result of substituting 3 for x in the polynomial of Ex. 1, Art. 8, viz.,

170, this being the remainder after division student can verify this by actual substitution.
is

by x -

3.

The

- 4 for x in Again, the result of substituting

is

- 855, as appears from Ex.

5,

Art. 8.

We saw in Art. 7 that

whose
as

as x receives a continuous series of values increasing from - oo to + GO , f(x) will pass through a corresponding continuous series. If

we

substitute in succession for x, in a polynomial

coef-

ficients are
.

given numbers, a

series of 0,

numbers such
2,
3,

..-6,-4,-3,-2,-l,

1,

4,

5,...,

and

calculate the corresponding values of /(), the process

the function.

may

be called the tabulation of

EXAMPLES.
1.

Tabulate the trinomial 2z2

x -

6,

for the following values of x

0, 1,

-4, -3, -2, -1,

Values of
x,

2,

3,

4.

14
series of

General Properties of Polynomials.

changes in the variable which it contains, it is plain that great advantage will be derived from any mode of representation which renders possible a rapid comparison with one another of the different values which the function may assume.

numerical

In the case where the function in question is a polynomial with coefficients, to any assumed value of x will correspond

one definite value of f(x).

We

proceed to explain a

mode

of

graphic representation by which it is possible to exhibit to the eye the several values of f(x) corresponding to the different values of x.

Let two right

(fig.

lines

OX,

OT
p.

1) cut one another at right

Q"

angles, and be produced indefinitely

in both directions.
called the axis of x

These

lines are
|

and

axis of y,

Lines, such as OA, respectively. measured on the axis of x at the

O A BC
r

lp

right-hand side of 0, are regarded as positive and those, such as

;

OA,
side,

measured at the left-hand

as negative.

to

OF which are above XX

it,

Lines parallel ', such as

such as

AP or &Q', are positive

and those below

Trigonometry.

are negative. These conventions are already familiar to the student acquainted with
as unity, may now be taken on or will be positive negative represented by a f : line measured on the series of numbers increasing from to + oo in the direction OX, and diminishing from to - oo in the

AT or AP,

Any

arbitrary length

OX

and any number

XX

direction OX'.

culate /(m)

Let any number m be represented by OA calA draw AP parallel to OF to represent f(m) in magnitude on the same scale as that on which OA represents m, and to represent by its position above or below the line OX
;
;

from

the sign oif(m).

represented

Corresponding to the different values of

P, Q,

by OA, OB, OC, &o., we shall have a series of points R, &c., which, when we suppose the series of values of

I ir

\:

in indefinitely

increased so as to include all numbers between

will trace out a continuous curved line.

co

and +

oo

This

curve will,

OX, The process here explained is also called tracing the function f(x). The student acquainted with analytic geometry will observe
that
is
it is

several points from the line exhibit to the eye the several values of the function f(x).

by

the distances of

its

equivalent to tracing the plane curve whose equation

application of this method it is well to begin points on the curve corresponding to certain

y =/(*). In the practical

by laying down the

small integral values of #, positive and negative. It will then in general be possible to draw through these points a curve which will exhibit the progress of the function, and give a general
idea of
its

character.

of course increase with the

The accuracy of the representation will number of points determined between

any two given values of the variable. When any portion of the curve between two proposed limits has to be examined with care, it will often be necessary to substitute values of the variable
separated

by

smaller intervals than unity.

The following

ex-

amples

will illustrate these principles.

EXAMPLES.
1
.

Trace the trinomial 2# + x

is

6.

unit of length taken in fig. 2. the line

The

one-sixth of

OD

1, Art. 9, the values of f(x) are given corresponding to the integral values 4 to + 4, inclusive. of x from

In Ex.

By means

of these

values

we

obtain
;

the positions of nine points on the curve seven of which, A, B, 0, D, E, F, G, are here represented, the other two correspond-

ing to values of f(x) which

limits of the figure.

lie

out of the

The student
the points

will find it a useful exercise

to trace the curve

more minutely between

JE in the figure, viz.

C and

by
''

calculating the values of/(#) corresponding to all values of x between 1 and 1 separated
as
is

'^-

by small

intervals, say of one-tenth,

done in the following examjil".

16
2.

General Properties of Polynomials.

Trace the polynomial

This
It

is

may

4 and 4. already tabulated in Art. 9 for values of x between be observed, as an exercise on Art. 4, that this function retains positive
all positive

values for

values of x greater than 2 '7, and negative values for

GO

all

values of x nearer to

than

2'7.

The
at all,

curve
cut

will, then, if it cuts the axis of

it at

a point (or points) corresponding to

(or values) of

some value

x between

2-7

and

2-7

so that if our object is to determine, or

to,

approximate

equation f(x) fined to the interval between

the positions of the roots of the 0, the tabulation may be con-

- 2-7

and

2*7.

This

is

a case in which the substitution of

integral values only of x gives very little help

towards the tracing of the curve, and where, consequently, smaller intervals have to be examined.

We
1,

tion for intervals

give the tabulation of the funcof one-tenth between the

1
;

integers

0; 0,

1, 2.

From

these values

the positions of the corresponding points on

the curve

may

be approximately ascertained,
fig. 3.

Y'
Fig. 3.

and the curve traced as in

Values of x
/(*)

-7
-15-96 -10-8 -6-46

-5 -4
2-24

-2 -1
5-04 5-72

Values of x
,

-1

4
4-32

/(*)

5-94

5-6

5-01

2-64

1-04

42

Values
/(*)

1-1

1-2

1-3

1-4

1-5

1-7

1-8

1-9

16

54

1-52

5-04

7-7 11-04 15-12

20

(a

The curve traced in Ex. 1 cuts the axis of a? in two points number equal to the degree of the polynomial) in other
: ;

words, there are two values of x for which the value of the given z polynomial is zero these are the roots of the equation 2x + x
6 =
0. viz.,

2,

and

1/5.

Similarly, the curve traced in Ex. 2

cuts the axis in three points, viz., the points corresponding to the 2 roots of the cubic equation 10# 3 - 17# + x + 6 = 0. The curve

Maxima and Minima

Values.

17

representing a given polynomial may not cut the axis of a: at all, or may cut it in a number of points less than the degree of the polynomial. Such cases correspond to the roots

imaginary

of equations, as will appear

more

fully in the next Chapter.

For

2 example, the curve which represents the polynomial 2^ + x + 2 will, when traced, lie entirely above the axis of x ; in fact, since

this function differs

from the function of Ex.

1 only

by the adis

dition of the constant quantity 8, each value of /(a?)

obtained
entire

by adding 8

to the previously calculated value,

and the

curve can be obtained by simply supposing the previously traced curve to be moved up parallel to the axis of y through a distance equal to 8 of the units. It is evident, by the solution of the
z equation 2x + x + 2 = 0, that the two values of x which render the polynomial zero are in this case imaginary. Whenever the

number

of points in which the curve cuts the axis of x falls short of the degree of the polynomial, it is customary to speak of the

curve as cutting the

11.

line in

imaginary points.

Haxima and Minima Values of Polynomials.

It

apparent from the considerations established in the pre- oo to + oo , ceding Articles, that as the variable x changes from
is

the function f(x) may undergo many variations. It may go on for a certain period increasing, and then, ceasing to increase, may commence to diminish ; it may then cease to diminish and

commence again
diminution

to increase

after

which another period of

may
At a

arrive, or the function

may

(as

in the last
in-

example
creasing.

of the preceding Art.)

go on then continually
is

stage where
to
;

the function ceases to

it it

increase

and commences

diminish,

said to

have attained a

maximum value
mences
value.
to

and when
it

ceases to diminish

increase,

is

said to have

attained a

and comminimum

polynomial may have several maxima, or several minima values, or both the number depending in general on the degree of the function. Nothing exhibits so well as a a maximum or graphic representation the occurrence of such
:

minimum

value

as well as the various fluctuations of

which

the values of a polynomial are susceptible. c

18

General Properties of Polynomials.

A knowledge

of the

maxima and minima

values of a func-

tion, giving the position of the points

where the curve bends

with reference to the axis, is often of great assistance in tracing It will be the curve corresponding to a given polynomial. shown in a subsequent chapter that the determination of these
points depends on the solution of an equation one degree lower than that of the given function.

CHAPTER

II.

GENERAL PROPERTIES OF EQUATIONS.

12.

THE

process of tracing the function /(#) explained in Art. 10

may be employed for the purpose of ascertaining approximately the real roots of a given numerical equation for when the cor;

responding curve is accurately traced, the real roots of the equation f(x) = can be obtained approximately by measuring the
distances from the origin of its points of intersection with the axis. With a view to the more accurate numerical solution of
this problem,

as well as the general discussion of equations

both numerical and algebraical, we proceed to establish in the

present Chapter the most important general properties of equations having reference to the existence, and number of the roots,

and the

distinction

between real and imaginary

:

roots.

By the aid of the following theorem the existence of a real root in an equation may often be established
Theorem.
the

If two real quantities a and b be substituted for

in

unknown quantity x

results having different signs, one

any polynomial f(x), and if they furnish plus and the other minus ; then the
least

equation f(x)

must have at

one real root intermediate

in

catno between a

and

b.

This theorem

is

an immediate consequence of the property

;

of the continuity of the function f(x) established in Art. 7 for since f(x) changes continuously from/() to/(), and therefore

passes through all the intermediate values, while x changes from a to b ; and since one of these quantities, /(a) or f(b), is positive,

and the other negative, it follows that for some value of x intermediate between a and &, f(x) must attain the value zero which is intermediate between /(a) and/().

c2

20
The

General Properties of Equations.

student will assist his conception of this theorem

by
is

reference to the graphic

method

of representation.

What

here proved, and what will appear obvious from the figure, is, that if there exist two points of the curved line representing the

polynomial on opposite sides of the axis 02", then the curve

It will joining these points must cut that axis at least once. also be evident from the figure that several values may exist

between a and
the axis.
tive

b for

which f(x) =

0,

i. e.

for

which the curve cuts

For example, in fig. 3, Art. 10, x = - 2 gives a negavalue (- 144), and x = 2 gives a positive value (20), and beexist three points of section

tween these points of the curve there

of the axis of x.f

Corollary. If there exist no real quantity ichich, substituted for x, makes f(x) = 0, thenf(x) must be positive for every real value
of x.

no value of

x= oc makes f(x) positive and it negative can make for if there were #, therefore, would the theorem the of this by equation any such value, Article have a real root, which is contrary to our present hypoFor
it is

evident (Art. 4) that

thesis.

With

reference to the graphic

mode

of representation

this theorem may be expressed by saying that when the equahas no real root, the curve representing the polytion /(#) = nomial f(x) must lie entirely above the axis of x.

13. Theorem. Every equation of an odd degree has at least one real root of a sign opposite to that of its last term.

This
Article.

is

an immediate consequence of the theorem in the Substitute in 'succession - <x> 0, GO for x in the
,

last

poly-

nomial/^).
for x

The
=0,

results are,
is

n being odd
;
t

(see Art. 4),

oo

f(x)

negative
is

x=

sign of /(#)

the same as that of a n

x=+
If a n
is positive,
i. e.

oo ,/(#) is positive.

and

0,

the equation must have a real root between - oo a real negative root ; and if a n is negative, the equa-

Imaginary Roots.
tion
root.

must have a

real root
is

between

and

oo

i.

e.

a real positive

The theorem

thus proved.
degree, whose last

14.

Theorem.

Every equation of an even

term

is

negative, has at least two real roots, one positive

and

the

other negative.

The

results of substituting

oo

0, oo are in this case

-GO,
0,

+,
-,

+ 00,
hence there
is

+;
oo

a real root between oo

;

and

0,

tween

and +

/.

e.

there exist at least one real negative,

and another beand

one real positive root. We have contented ourselves in both this and the preceding Articles with proving the existence of roots, and for this purpose
it is sufficient

as

to substitute very large positive or negative values, done, for x. It is of course possible to narrow the limits within which the roots lie by the aid of the theorem of

we have

Art. (4), and still more by the aid of the theorems respecting the limits of the roots to be given in a subsequent Chapter.
15.

Existence of a Root in the General Equation.

Imaginary Roots.
real root

We
of

have now proved the existence of a

every equation except one of an even

in

the

case

degree whose
real root at

last term is positive. Such an equation may have no

It is necessary all. then to examine whether, in the

absence of real values, there may not be values involving the ima-

ginary expression */-

1,

which,

when

substituted for x, reduce the

;

polynomial to zero or whether there may not be in certain cases

both real and imaginary values
of the variable
Fig. 4.

which

satisfy the equation.

We

take a simple

22
example

General Properties of Equations.

to illustrate the occurrence of such

imaginary

roots.

As

already remarked (Art. 10), the curve corresponding to the

/(a?)

polynomial
3 2a? + x + 2

lies

entirely above the axis of #, as in

= f(x)

has no real roots

but

it

The equation has the two imaginary roots

fig.

4.

as

is

evident

by the

solution of the quadratic.

We

observe,

therefore, that in the absence of

this case

any two imaginary expressions which reduce the polynomial

real values there are in

to zero.

The general
tion
is,

proposition of which this furnishes an illustrathat Every rational integral equation

a^x
n

+ a xn
v

~
l

+ a z x n ~l+

+ a n .iX + a n =

must have a root of

the form

a and

($

being real finite quantities.

This statement includes

both real and imaginary roots, the former corresponding to the = 0. value
/3

the proof of this proposition involves principles which could not conveniently have been introduced hitherto, and which will present themselves more naturally for discussion
in subsequent parts of the work,

As

we

defer the demonstration

until these principles have been established.

therefore,

For the

present,

we assume

the proposition, and proceed to derive

certain consequences
16.

from

it.

Theorem.

Every equation of n dimensions has n

roots,

and no more.

We first observe that if any quantity h is a root of the

tion/^) =
This
is

equaa root

0,

then/(#)

is

divisible
;

by x-h without a remainder.

=
0,
i. e.

evident from Art. 9

for if /(/*)

if

is

=0, ,Rmustbe=0.

Imaginary Roots.
Let, now, the given equation be
n /(*) s x

23

+p

xn

~l

+ p,xn

~z

+ ...

+p n.

+p n

0.

This equation must have a root, real or imaginary (Art. 15), which we shall denote by the symbol a,. Let the quotient, when we have then the identical f(x) is divided by x eti, be ^ (x)
l ;

equation

/()-(-!> ft ();
Again, the equation fa (x) = 0, which is of n - 1 dimensions, must have a root, which we represent by a 2 Let the quotient obtained x - a 2 be Hence
.

by dividing

fa (a)

by

fa \x)

*(*).*(*-*)#(*)
and
where
fa (x) is of
/.

f(x)

^(x-

ai

)(x-

a 2 ) fa (x) ,

n - 2 dimensions.

Proceeding in this manner, we prove that f(x) consists of the product of n factors, each containing x in the first degree, and a n numerical factor n (x) Comparing the coefficients of # , it is
.

(f>

plain that fa

(x)

1.

Thus we prove the

identical equation
l

/(0)(0-Oi)(0-ab)(0-)
It
ties
cti,

..... (x-a n. )(x-a n ).

is

a2,

evident that the substitution of any one of the quantia n for x in the right-hand member of this equation
.

will reduce that to zero

;

member
. .
.

to zero,

and

will therefore reduce f(x)

that
<n,

is
,

to say, the equation f(x) a n _i, a n

.

has for roots the n

;

quantities
for if

a2 as

And

it

can have no other roots

.
.

any quantity other than one

of the quantities a 1? a 2 ,

an

be substituted in the right-hand member of the above equation, the factors will be all different from, zero, and therefore the product cannot vanish.

Corollary.
(to

Two polynomials of the n th

degree cannot be equal

one another for more than n values of the variable without bcimj

completely identical.

For
tion of

if

their difference be equated to zero,

t^^^^egree,

which can be

satisfied

we obtain an equaby n values only of

the variable^SHikeach coefficient be separately equal to zero.

24

General Properties of Equations.

The theorem of this Article, although of no assistance in the solution of the equation f(x) = 0, enables us to solve completely the converse problem, i. e. to find the equation whose roots are
any n given quantities. The required equation is obtained by multiplying together the n simple factors formed by subtracting from x each of the given roots. By the aid of the present

when any (one or more) of the roots of a given are known, the equation containing the remaining equation roots may be obtained. For this purpose it is only necessary
theorem
also,

to divide the given equation

by the product
factors.

of the given bino-

mial factors.

The

quotient will be the required polynomial

composed of the remaining

EXAMPLES.
1.

Find the equation whose roots are

-3,
2.

-1,

4,

5.

Am.
The equation
x * _ 6*3+ 8.*"- 17*

s*

- 5s 8 - 13s3 + 53#+

60

= 0.

10

has a root 5

find the equation containing the remaining roots.

division of Art. 8.

Use the method of

Ans. x* - x 1 +
3.

2>x

-2=

0.

Solve the equation

a;

-16a; 3 + 86*2 -l76*

+105 = 0,
3, 5.

two

roots being 1

and

7.

Ans. The other two roots are

4.

Form

the equation whose roots are

3
1

-2'
5.

3'
7-

Ans.
Solve the cubic equation

Hz 3 - 23z2 - 60z + 9 = 0.

Here

it is

evident that x

1 satisfies

the equation.

Divide by x

1,

and solve the

resulting quadratic.

The two

roots are

found to be

.i+iy-a,
6.

__'^.

Form an

equation with rational coefficients which shall have for a root the

irrational expression

Equal Roots.

25
-5

This expression has four different values according to the'Iitferent combinatk of the radical signs, viz.

\/P-V<l,
The required equation
(x
is,

-r

therefore,

- </p -\/q)

(x +

*/P

vA) (x ~ VP + V'l}

(*

VP ~ VI} =

0,

or, finally,

z4 - 2 (p

-f

?) X*

(p

- ?) 2 = 0.
factors

17.

Equal Roots.

It

must be observed that the n

of which a polynomial f(x] consists need not be all different from one another. The factor x - a, for example, may occur in

the second, or any higher power not superior to n. In this case is still said to have n roots, two or more the equation f(x) =

being

now

equal to one another

:

and the

root a

is

called a

mul-

is repeated. reference to the graphic construction in Art. 10 will help to explain the occurrence of multiple roots.

tiple root of the equation number of times the factor

double, triple, &c., according to the

(fig.

3)

We

see

by an

inspection of the figure that the two positive roots of the 3 are nearly equal, and we may 17 x2 + x + 6 = equation 10^ conceive that a slight addition to the absolute term of this poly-

nomial, which
parallel
effect of

is,

as already explained, equivalent to a small of the whole curve,

movement upwards

would have the

rendering equal the roots of the equation thus altered. would no longer cut the curve in two In that case the line

OX

distinct points, but

curve

it is

would touch it. Now, when a line touches a properly said to meet the curve, not once, but in two

coincident points.

the theory of plane curves will have no difficulty in illustrating in a similar manner the occurrence of a triple or higher multiple root.

The student acquainted with

form the connecting link between real and We have just seen that a small change in the imaginary form of a polynomial may convert it from one having real roots into another in which two of the real roots become equal. A further small change may convert it into a form in which the two

Equal

roots

roots.

26
roots

Genera^Properties of Equations.
become in&ginary.
is

Let us suppose that the above poly-

nomial

further altered

lute term.

by another small addition

to the abso-

the axis

/X

then have a graphic representation in which cuts the curve in only one real point, viz., that core shall

root, the two points of section corretwo roots having now disappeared. positive sponding 3 17#2 + x + 28, Consider, for example, the polynomial 10# which is obtained from that of Ex. 2, Art. 10, by the' addition

respor/ng to the negative

to the

of 22.

The student can

easily construct the figure

in fig. 3 will now lie much corresponding to Divide by x + 1, and obtain the trinomial 10#- - 27'x + 28 which

the point above the axis of x.

:

contains the remaining two roots.

They
27

are easily found to be

"

lj

v/391 /

20"

-^0^~

We
that
root to

when a change

as- well as in the example of Art. 15, form of the polynomial causes one real disappear, a second also disappears at the same time, and

observe in this case,

of

the two are replaced by a pair of imaginary roots. The reason of this will be apparent from tjie proposition of the following
Article.

Imaginary Roots enter Equations in Pairs. The proposition to be now proved may be stated as follows
18.

If an equation f(x) =

0, ichose coefficients are all real quantities,

it

have for a root the imaginary expression a + j3 v/- 1, have for a root the conjugate imaginary expression a have the following identity

must
1-

also

fly

We

this identity,,

Let the polynomial f(x) be divided by the second member of and if possible let there be a remainder Rx + R'.

We have then the identical equation

where

is

the quotient, of n - 2 dimensions in

x.

Substitute in

Roots.
this identity a

27

j3

v/-

1 for x.

This,
3

by

to vanish.

It also causes (x

a)

/3

hypothesis, causes /(*) to vanish. Hence

From

this

we

obtain the two equations

since the real

and imaginary parts cannot destroy one another

hence

Thus

the remainder

Rx + R' vanishes

and, therefore, /(.r)

is

divisible without

remainder by the product of the two factors

The equation
as the root a

has, consequently, the root a-j3'<v/- 1 as well

1.

+ ft v/total

Thus the
with real

number

of imaginary roots in an equation

;

always even and every polynomial may be regarded as composed of real factors, each pair of imaginary roots producing a real quadratic factor, and each real root procoefficients is

ducing a real simple factor. The actual resolution of the polynomial into these factors constitutes the complete solution of the
equation.

We

observed in Art. 17 that equal roots

may

be considered

as the connecting link between real and imaginary roots. This statement may now be regarded from another point of view. 2 Suppose a polynomial has the quadratic factor (#-a) + , and
let its

form be altered by means of slight alterations in the value

of k.
roots
;

When k is negative,
when
k =
0,

the quadratic factor gives a pair of real this factor has two equal roots, a ; when k is

positive, the factor

A proof exactly
roots,

has two imaginary roots. similar to that above given shows that nun/

of the form a

y/y, enter equations whose

coefficient*

arc

rational in pair*.

28

General Properties of Equations.

EXAMPLES.
1.

Form

a rational cubic equation which shall have for roots

1,

Ans. * 3 -7
2.

Form

a rational equation which shall have for two of

its

roots

Ans. z*-12x* + l2x*-3l2x

3.

+ 676 = 0.

Solve the equation

*4 +

2^-5*2 + 6z + 2 = 0,

which has a root

Ans. The roots are
4.

-2 + \/ 3,

Solve the equation

one root being

2

V^T.
-4s. The roots are
2,

19.
rule,

Descartes'

Rule of Signs
us,

Positive Roots.

This

which enables

by the mere
:

inspection of a given equa-

tion, to assign a superior limit to the

roots,

number
+

of its positive

may

be enunciated as follows
member.

No

equation can have wore

to -,

positive roots than it has changes of sign from

+, in the terms of its first

and from -

to

We
which

shall content ourselves for the present with the proof

usually given, and which is rather a verification than a general demonstration of this celebrated theorem of Descartes. It will be subsequently shown that the rule just enunciated, and other similar rules which were discovered by early investigators
is

number of the positive, negative, and imaginary roots of equations, are immediate deductions from the more general theorems of Budan and Fourier.
relative to the

Let the signs of a polynomial taken

other in the following order
:

at

random succeed each

+ + - +

---

+ + - +

-.

In

this there are in all seven

-,

changes of sign, including

It
is

changes from + to

and from -

to +.

proposed to show

Descartes' Rule of
that
if this

Sit//

29

polynomial be multiplied by a binomial whose signs, corresponding to a positive root, are + -, the resulting poly-

nomial will have at

original.

least

one more change of sigh than the

We write down
as follows
:

only the signs which occur in the operation

.+

---

+ + - + H

+ - + +

--

- + - +

Here in the third line the ambiguous sign is placed wherever there are two terms with different signs to be added. observe in this case, and it will readily appear also for every

We

other arrangement, that the effect of the process is to introduce the ambiguous sign wherever the sign + follows -f, or - follows -, in the original polynomial. The number of variations of sign

never diminished. There is, moreover, always one variation added at the end. This is obvious in the above instance, where the
is

original polynomial terminates with a variation if it terminate with a continuation of sign, it will equally appear that the cor;

responding ambiguity in the resulting polynomial must furnish one additional variation either with the preceding or with the

Thus, in even the most unfavourable case that, namely, in which the continuations of sign in the original remain continuations in the resulting polynomial, there is one
superadded sign.
:

variation added
effect

and we

may

of the

multiplication of a -polynomial

conclude in general that the by a binomial

factor x
sign.

- a

is

to introduce at least one additional

change of

factors corresponding to the negative

;

Suppose now a polynomial formed of the product of the and imaginary roots of an

equation the effect of multiplying this by each of the factors x - a, x - /3, x - y, &c., corresponding to the positive roots a, |3, 7, &c., is to introduce at least one change of sign for
each; so that

when

the complete product

is

formed containing

30
all

General Properties of Equations.

the roots,

we

least as

many

conclude that the resulting polynomial has at changes of sign as it has positive roots. This is

Descartes' proposition. 20. Descartes' Rule

of Signs

Negative Roots.

In

order to give the most advantageous statement to Descartes' rule in the case of negative roots, we first prove that if - x be substi-

tuted for x in the equation /(#) = 0, the resulting equation will have the same roots as the original except that their signs will

be changed.

This follows from the identical equation of Art. 16

= (x- aO (x - a 2) (x - a 3 )
.

f(x)

(x

a),

from which we derive

From

this it is evident that the roots

of/(-#) =
d

are

Hence the negative roots of/(#) are we may enunciate Descartes' rule
lows
:

positive roots of /(-#),

and
roots

for negative roots as fol-

No

equation can have a greater

number of negative

than there are changes of sign in the terms of the polynomial f(-x). 21. Use of Descartes' Rule in proving the existence

of Imaginary Roots.

It is often possible to detect the existence of imaginary roots in equations by the application of Descartes' rule ; for ifjt should happen that tha ^nm jrgjfJTg greatest possible
possible

number

of positivejroots,added to the greatest

number^oFnegative roots^jsjess than the degree of the are suTe^fjbhe^xistence^f imaginary roots. Take, equatiojnjjjge
for example, the equation

one positive

This equation, having only one variation, cannot have more than root. Now, changing x into - #, we get

and since this has only one variation, the original equation cannot have more than one negative root. Hence, in the proposed

Theorem.

;;l

equation there cannot exist more than two real roots. It lias, This application of therefore, at least six imaginary roots.
Descartes' rule
is
it is

available

only in the case of incomplete

equations

for

easily seen that the

sum

of the

number

of

variations in f(x) and/(-rr) is exactly equal to the degree of

the equation
22.

when

it is

Theorem.

//'

complete. two numbers a and

b,

substituted for

in

the polynomial f(x), give results with contrary signs, an odd

number

lies between them; and of real roots of the equation f(x] = if they the same with sign, either no real root or an even numgive results

ber of real roots

lies

between them.

This proposition, of which the theorem in Art. 12

ticul

wh
ar*

'

&
u. S

is a parcontains in the most general form the conclusions Sa drawn as to the roots of an equation from the
^

g i oots are

^7
:

^ rst member when two given numbers

1 for x.
^ion

the pi.

proceed to prove the first part of the second part is proved in a precisely similar

We

manner.

Let the following the equation f(x) =

roots ai, a 2 ,

a^,,

and no

others, of

lie between the quantities a

and

b,

of

which, as usual, we take a to be the Let (x) be the quotient when/(#)

of the

lesser.
is

divided

by the product
have, then,

factors (x-a^)

(#-a

2)

(x-a m ).

We

the identical equation

= Putting in this successively x

f(a)
i

a,

x=

b,

we

obtain
,

= (a-a )(a-a 2 ) .... (a - a m ) $ (a) /(fl)-(ft-a,)(6-a 8 ) .... (b-a m )t(b).

have the same sign for if they had different signs there would be, by Art. 12, one root at least of the between them. By hypothesis, /(a) and/(&) equation $ (x) =
<j)(a)

Now

and

(b)

have different signs

hence the signs of the products

(a-ai)(a-a z ) .... (a-a m ), (b- ai )(b-a z ) .... (b-a m ),

32
are different
its
;

General Properties of Equations.

but the sign of the second is positive, since all hence the sign of the first is negative ; factors are positive
;

but

all

the factors of the

;

first

are negative

therefore their

number must be odd

which proves the proposition.

In this proposition it is to be understood that multiple roots are counted a number of times equal to the degree of their
multiplicity.

It is instructive to apply the graphic method of treatment to the theorem of the present Article. From this point of view it appears almost intuitively true for it is evident that when any
;

two points are connected by a curve, the portion of the curve between these points must cut the axis an odd number of times when the points are on opposite sides of the axis and an even
;

number

of times, or not at

all,

when

the points are on tb

ime

side of the axis.

EXAMPLES.
1.

If the signs of the terms of

an equation he

all positive,

it

cannot have a

positive root.
2. If the signs of the terms of any complete equation he alternately positive and negative, it cannot have a negative root. 3. If an equation consist of a number of terms connected hy + signs followed hy a number of terms connected by signs, it has one positive root and no more.

Apply Art. 12, substituting and oo and Art. 19. 4. If an equation involve only even powers of #, and
;

if all

the coefficients have

positive signs,

it

cannot have a real root.

Apply Arts. 19 and 20. 5. If an equation involve only odd powers of x, and if the positive signs, it has the root zero and no other real root.

coefficients

have

all

6. If an equation be complete, the number of continuations of sign in f(x) is the same as the number of variations of sign in/( x). 7. "When an equation is complete if all its roots are real, the number of positive
;

equal to the number of variations, and the number of negative roots .to the number of continuations of sign.
roots
is

is

equal

8. An equation having an even number of variations of sign must have its last sign positive, and one having an odd number of variations must have its last sign

negative.

Take the highest power of x with positive coefficient (see Art. 4). 9. Hence prove that if an equation has an even number of variations it must have an equal or less even number of positive roots and if it has an odd number of
;

variations

it

must have an equal or

less

odd number of positive roots

in other

Examples.
words, the number of positive roots
differ

33

when

less

than the number of variations nm-t

by an even number. and oo and apply Substitute

it
,

from

Art. 22.

10.

Find an

inferior limit to the

number of imaginary

roots of the equation

Ans. At
1 1
.

least

two imaginary

roots.

Find the nature of the

roots of the equation

z4

15s2 + 7.T- 11

= 0.
one negative, two imaginary.

Apply
12.

Arts. 14, 19, 20.

Ans. One

positive,

Show

that the equation

x*

+ qx + r = 0,
one negative and two imaginary
roots.

where q and r are essentially

13.

positive, has

Show

that the equation

x*-qx + r = 0,
where q and r are
roots are either
14.

essentially positive, has one negative root

and that the other two

imaginary or both positive.

Show

that the equation

& C A xa xb xc
z
2

\/

xl
It

= x-m

where

a, b, c,

....

are

numbers
and a

all different

from one another, cannot have an

imaginary

root.

Substitute a

+ ft \/

ft

in succession for #,
ft

and

subtract.

\Ve

get an expression which can vanish only on the supposition 15. Show that the equation

= 0.

x* has,
is

=
1
,

when n

is

even,

odd, the real root

1,

two real roots, and no other

and

and no other

real root

and,

when n

real root.

This and the next example follow readily from Ails. 19 and 20.
16.

Show

that the equation

xn
;

+1 =
when n
is

has,

when n

is

even, no

real root

and,

odd, the real root

1,

and no

other real root.

17. Solve the equation

z 4 + 2qx* + 3? 2 z? + 2q*x - r 4 = 0.
This
is

equivalent to

The

different signs of the radicals give four combinations,

and the

ex;

here written involves the four roots.

34
18.

General Properties of Equations.

Form
the equation which has for roots the different values of the expression

2+0X/7+V/11 +
where
If
2

\/7,

=1.
restriction

no

have 8 values.
in

had been made by the introduction of 0, this expression would The \/1 must now be taken with the same sign where it occurs
it.

under the second radical and free from

all.

There

are, therefore,

only four values

Ans. z 4 - 8z3 -12z 2 +84z-63

= 0.

19.

Form

the equation which has for roots the four values of

where
20.

1.

-9 + v/ 137 + 3v34- 20 Vl37, Ans. x* + 36z 3 - 400# 2 - 31 680; + 7744 = 0.

all

equation with rational coefficients which shall have for roots values of the expression
0i

Form an

the

/p + h V
2

<1

+ h ^/r,
3 3

where
There are eight

0i

= l,

1,

=1.

different values of this expression, viz.,

-VP

Assume
*=
Squaring,
0i

VP +
3

fa

Vq +

fa

*/*

we have
# 2 =p

+q+r+ 2

(0 2

i/gr+ fa 0i */rp + 0i

Vpq]

Transposing, and squaring again,

'

Transposing, substituting x for

Oj.

\/'p +

\/ q +

^/r and squaring, we obtain

t

the final equation free from radicals

{z
This
is

- 2#2 (p + q + r) + p + ? 2 + r z - 2qr - 2rp z 2pq} = Qtyqrx*.

ten.

an equation of the eighth degree, whose roots are the values above writQ\, 02, 0s have disappeared, it is indifferent which of the eight roots *\/P+\Sg\/r is assumed equal to # in the first instance. The final equation
Since

is

that which would have been obtained if each of the 8 roots had been subtracted from x and the continued product formed, as in Ex. 6, Art. 16.
y

CHAPTER

III.

RELATIONS BETWEEN THE ROOTS AND COEFFICIENTS OF EQUATIONS, WITH APPLICATIONS TO SYMMETRIC FUNCTIONS OF THE
ROOTS.
23.

Relations between the Roots and Coefficients.

for simplicity the coefficient of the highest

Taking
tion

as unity,

and representing,
.
.

as in Art. 16, the n roots of

power of x an equa:

by

ai, ao, as,

a,,,

we have

the following identity

Xn + piXn
= (xai) (x

a a) (x - a 3)

(x

- an ).

(1)

When

the factors of the second

member

of this identity are

multiplied together, the highest power of x in the product is a? ; n~ l is the sum of the n quantities - QI, - a 3 &c., the coefficient of x
,

viz.,
is

z the roots with their signs changed ; the coefficient of x"the sum of the products of these quantities taken two by

two; the
three
all

coefficient of

xn

~3

is

the

sum

of their products taken

by

three

and

so on, the last

term being the product of

the roots with their signs changed. Equating, therefore, the coefficients of x on each side of the identity (1), we have the

following series of equations

(eti

a2 +

ai a 3

-f

a2 a3

+....+
...

a,i_ 2 a,,_i a,,),

(2)

n p n = (- l) ai a 3 a 3

a-i

which enable us to

state the relations

:

between the roots and

coefficients as follows

D2

36

Roots and

Coefficients

of Equations.
equation,
the coefficient

Theorem.
whose highest term
its

In every
is is

algebraic

of

unity, the coefficient p^ of the second

term with

equal to the sum of the roots. coefficient p z of the third term is equal to the sum of the products of the roots taken two by two. The coefficient p 3 of the fourth term with its sign changed is
sign changed

The

equal to the

sum of the products of the

roots taken three by three

and so

on, the signs of the coefficients being taken alternately negative

and positive, and the number of roots multiplied together

in each term
till

of the corresponding function of the roots increasing by unity,

finally that function is reached which consists of the product of the

roots.

When

n the coefficient a of x

is

not unity (see Art.

1),

we

must divide each term

roots is then equal to

of the equation
*
;

by

it.

The sum

of the

the

sum

of their products in pairs is

#0

equal to a
Cor. 1. lute

and
"

so on.

term of
Cor. 2.

Every root of an equation is a divisor of the abso-/ the equation/ If the roots of an equation be all positive, the coef-

^^f*^

highest power of x) will be alternately positive and negative ; and if the roots be all negative, the coefficients will be all positive. This is obvious from the
ficients (including that of the

equations
24.

(2) [cf.

Arts. 19

and 20].
Since the equations

Applications of the Theorem.

(2) of the preceding Article supply n distinct relations between the n roots and the coefficients, it might perhaps be supposed

that

thereby gained in the general solution Such, however, is not the case for suppose it were attempted to determine by means of these equations a root, di, of the original equation, this could be effected only by the
is

some advantage

of the equation.

by means of the given equations, and the consequent determination of a final equation of which
elimination of the other roots
ai is

one of the
is

roots.
it

Now,

in whatever
for solution

way

this final equaa,,

tion

obtained,

must have

not only

but each

Theorem.
of the other roots a 2 , a 3 ,
.

37
the roots enter in

an

for, since all

in the equations (2), if it had been proposed to determine a> (or any other root) by the elimination of the rest,

the same

manner

our final equation could differ from that obtained for a! only by the substitution of a 2 (or that other root) for ai. The final equation arrived at, therefore, by the process of elimination must

have the

ti

quantities

ai,

a 2,

.....

a n for roots

and cannot,

consequently, be easier of solution than the given equation. This final equation is, in fact, the original equation itself, with the root we are seeking substituted for x. This we shall show
for the particular case of a cubic.
is

The

general, and may be applied to an equation Let a, /3, 7 be the roots of the equation x3 +piar +p 2 x +p 3 = 0.

process here employed of any degree.

We have,

by Art.

23,

lh
Pz

= -(a + /3 + y),
=
a/3

+ ay +

]3y,

^ 3 = -a/3y.
Multiplying the
first

of these equations

by

[a

the second

by
or

a,

and adding the

three,

we

find

which,

is

the given cubic with a in the place of

as

x.

The student can take

an exercise

to prove the

same

result

In the correin the case of an equation of the fourth degree. successive the case the of equations of general sponding treatment
Art. 23 are to be multiplied

by

a"-

a"'

3
,

a"'

Although the equations

(2) afford, as

and added. we have just seen, no

,

&c.,

assistance in the general solution of the equation, they are often of use in facilitating the solution of numerical equations when

any

particular relations
also

among

the roots are

known

to exist.

They may

be employed
relations

to establish the relations

which

must obtain among the

responding to

coefficients of algebraical equations cor-

known

among

the roots.

88

Roots and

Coefficients

of Equations.

EXAMPLES.
1.

Solve the equation

z3 - 5*2 -16# + 80 = 0,
the

sum

of

two of

its

roots being equal to nothing.

Let the roots be

a, 0, 7.

We have then
5,-

a0

*7

+ 07 = -16,

Taking

3+7=0, we
we
obtain

have, from the

first

of these,

o= 5

and from either the

.

second or third

0y
4.

- 16

We find for & and 7 the values 4 and - 4

Thus

the threejroots are 5, 4, 2. Solve the equation

two

of its roots being equal.

Let the three roots be

o, a, 0.

We have
a2

+ 2a0=0,
The
roots are 2, 2,

from which
3.

we

find a

= 2, and

=-

1.

1.

The equation

has two pairs of equal roots Let the roots be o, o, 0,

find them.
;

ft

we

have, therefore,

from which we obtain for a and

4.

/8

the values 1 and

3.

Solve the equation

x 3 -9^ 3 +H;e + 24 = 0,

two of whose
easily obtain

roots are in the ratio of 3 to 2.

a, 0, 7,

Let the roots be

with the relation 2 a =

3)8.

By

elimination of a

we

50 + 27 =

18,

from which we have the following quadratic

190
2

for

)8

-900 +56 = 0.

Examples.
The
1.

roots of this are 4,

and

the former gives for a and

7 the
what

values 6 and
is

The

three roots are 6, 4,

of

1.

The student

will here ask

the signi-

ficance of the value

&

and the same

difficulty

may have

presented itself in

the previous examples.

It will he observed that in examples of this nature

we

never require all the relations hetween the roots and coefficients in order to determine the required unknown, quantities. The reason of this is, that the given condition establishes one or more relations amongst the roots. "Whenever the equations

more than one system of values for the roots, the actual by the condition that they must satisfy the equation (or equations) between the roots and coefficients which we have not made use of in
employed appear
to furnish

roots are easily determined

determining them. Thus, in the present example, the value satisfying the omitted equation

=4

gives a system

while the value

to be rejected.
5.

=
iy

gives a system not satisfying this equation, and

is

therefore

Solve the equation

whose

roots are in arithmetical progression.

5,

Let the roots be a -

o,

we have

at once

3o=
3cr-8 =
2

9,

23,

from which we obtain the three roots

6.

1, 3, 5.

Solve the equation

=
whose roots are in arithmetical progression.

0,

Assume
7.

for the roots

35,

5,

a+

5,

+ 38.
Ans. -5, -2,
1,

4.

Solve the equation

27s3 + 42s2

- 28* - 8 = 0,

whose

roots are in geometric progression.

Assume
have o3 =

for the roots ap, a, -. From the third of the equations (2), Art P Q 2 = or a -. Either of the remaining two equations gives a quadratic for p. ,

8.

Solve the equation

3*4

- 40* 3

-I-

130*2 - 120*

+ 27 =

0,

40
whose

Roots and

Coefficients

of Equations.

roots are in geometric progression.

Assume

for the roots

-,

ap, op

Employ the second and fourth

Ans. i
o

of the equa-

tions (2), Art. 23.

9.

.1,

3,

9.

Solve the equation

x*

+ 15s3 + 70s2 + 120s +

64 =

0,

whose
10.

roots are in geometric progression.

Ans.

1,

2,

-4, -8.

Solve the equation

6s 3

-lls2 + 6s-l = 0,

whose

roots are in

Take the

roots to be o,

harmonic progression. We have here the relation ft, y.

1

1_2
&c.

hence

&y + ya + aft = 3ya

Ans.
11.

1,

-,

-.

Solve the equation

81s3 -18s 2 -36s + 8

= 0,

whose

roots are in

harmonic progression.

12.

If the roots of the equation

be in harmonic progression, show that the mean root

13.

is
.

g.

The equation

:-21 =

has two roots equal in magnitude and opposite in sign ; determine all the roots. Take a + = 0, and employ the first and third of equations (2), Art. 23.

Ans.
14.

v/3, -v/3,

l+-v/~6.

The equation
3z 4 - 25s 3 + 50s 2 - 50s + 12

has two roots whose product

is

find all the roots.

Ans.
15.

6,

-,

1\/-!-

One

of the roots of the cubic

s3
is

px~ + qx

r=

double another

show

that

it

may

be found from a quadratic equation.

Examples.
16.

Show

that all the roots of the equation

can be obtained when they are in arithmetical progression. Let the roots be o, o + 5, a + 25, a + (n - 1) 5. The
.
.

first

of equations (2)

gives

w(n-l) _!__;

(1)

Again, since the sum of the squares of any number of quantities is equal to the square -of their sum, minus twice the sum of their products in pairs, we have the
equation

pr - 2/? 2 =

+ (a +

S)

+ (a + 25) 2 +

(2)

from n times the equation (2), \ve find S 2 in terms Thus all the roots can be (1). expressed in terms of the coefficients p\ and p%. 17. Find the condition which must be satisfied by the coefficients of the equaSubtracting the square of
(1)

of pi

and

p-z.

We

can then find a from equation

tion

x3

pxz
by
a relation a

when two
18.

of its roots

o,

are connected

+ ft = 0.
Ans, pq
r

= Q.

Find the condition that the cubic

#3 -px"~ -f qx
should have
19.
its

r=0
Ans. p 3 r
q
3

roots in geometric progression.

0.

Find the condition that the same cubic should have its roots in harmonic " Ans. 2lr z - 9pqr + 2q 3 = 0. progression (see Ex. 12).
20.

Find the condition that the equation

should have two roots connected by the relation a

case

-f

and determine in that

ft
;

two quadratic equations which

shall

have for roots

(l)px~

(1) o,

and
x* +

(2) 7, 8.

Ans. pqr-p-s-r^^O,

+ r = 0,

(2)

px+^ = 0.

21. Find the condition that the biquadratic of Ex. 20 should have its roots conAns. p 3 nected by the relation + 7 = o + 5. pq + 8r = 0. 22. Find th<; condition that the roots o, )8, 7, 5 of

x*

+px 3

+ qx-

+ rx + s =
Ans. p 2 s - r2 =
is satisfied

should be connected by the relation o)3 = yd. 23. Show that the condition obtained in Ex. 22
the biquadratic are in geometric progression.

0.

when

the roots of

42
25.

Roots and

Coefficients

of Equations.

exists

Depression of an Equation when a relation between two of its Roots. The examples given

in the preceding Article illustrate the use of the equations connecting the roots and coefficients in determining the roots in particular cases when known relations exist among them. in general, that if a relation of the form /3 = 0(a) exist between two of the roots of an equation f(x) = 0, the equation
shall

We

now show

may

be depressed two dimensions.

<j>

Let

(x)

be substituted for x in the identity

n f(x) = a Q x

then

/(0

(x)

s a (0 (x) )'* + a

n~ l
1

(<j>(x))

+ ---- + a n -i $ (x) + an

represent, for convenience, the second member of this identity by F(x]. Substituting a for x, we have

We

^) =/(*()) =/(0)-<>;
hence a -satisfies the equation JF(a?)*=0, and it also satisfies the = hence the polynomials/^) and F(x] have a equation / (x) common measure x - a thus a can be determined, and from it
; ;

0(o) or
sions.

]3,

and the given equation can be depressed two dimen-

EXAMPLES.
1.

The equation
x z - 5^2

x + 20

has two roots whose difference

=3
the

find them.

Here
it

- a = 3,
=5
;

=
1x

4-

substitute

x+

3 for

becomes x ^ + 4#2

10

common measure
is

x in the given polynomial f(x) of this and/(#) is x - 2 from

; ;

which
2.

a= 2,

the third root

- 2. _ 13
2/3 -f
_j_

The equation

=o
find all the roots.

has two roots connected by the relation

3a

=7

^*.
It

1,

2,

l\/^.

may

and F(x) have common by the ordinary process

be observed here, that when two polynomials /(#) factors, these factors may be obtained
of finding the

common

measure.

Thus,

The Cube Hoots of Unity.

if

43

we know that two given equations have common roots, we can obtain these roots by equating to zero the greatest common measure of the given polynomials.
EXAMPLES.
1.

The equations

2^ + 5^- 6xS*3

= 0,

+7z-- 11* -15 =

Ans. -1,

have two common roots

2.

find them.

-3.

The equations

+r

=0,

have two common roots

the third root of each.

find the quadratic

whose

roots are these two,

and

find also

Ans

P~P

r^Lo, -

-r(p^P)
'

-r(p-p')
'

26.

The Cube Roots of

Unity.

Equations of the

forms

consisting of the highest and absolute terms only, are called The roots of the former are called the n n th binomial equations.
roots of unity.

A general discussion of these forms will be given

in a subsequent Chapter. "We confine ourselves at present to the simple case of the binomial cubic, for which certain useful
It has been properties of the roots can be easily established. Art. Ex. that the shown roots of the cubic 5, already (see 16),
z*

- 1 =
1
1

are

1,

-o + ~v

/-o

-3,

-?j-rv~3.

other

If either of the imaginary roots be represented by w, the 2 is easily seen to be w by actually squaring or we may
, ;

see the

same thing

as follows
;

If
3

tu

be a root of the cubic,

1,

<o

must

also

be a root

for, since

w =

we

get,

by squaring,

44
6

Roots and
2 3

Coefficients

of Equations.

a)

x*

= 1, or - 1 = 0.

(o>

We

thus showing that have then the identity

1,

w2

satisfies

the cubic

Changing x

into
z

#,

we
(x

get the following identity also

x + 1 =

1) (x

+ w) (x +

o>

),

which furnishes the roots of

^ + 1 = 0.
in any product of quantities involving the imagicube roots of unity any power higher than the second nary 2 presents itself, it can be replaced by w, or w or by unity for
, ;

Whenever

example,
<o
4

= o> 5
first

GJ

= w,

w 5 = to 3

2
.

o>

= w2

<u

o>

3
.

o>

1,

&C.

or second of equations (2), Art. 23, gives the following property of the imaginary cube roots
:

The

+w+w =
2

0.

the aid of this equation any expression involving real quantities and the imaginary cube roots can be written in either

By

of the forms

P + w Q, P +

o>

Q.

EXAMPLES.
1.

Show

that the product

(cam

+ u-n)
:

2
(o>

w -f tan)
Ans. m~
mn-\-n~.

is rational.

2.

Prove the following

identities

m 3 + n? = (m + n) ((am + m _ W 3 = w _ n (<am
z r
)

o>

w)
n)

w + (an),
(an).

(a

(uPm

3.

Show

that the product

(a

+ (aft + or 7) (a+

+ (ay)
Ans. a2 +/8 2 + 72

is rational.

- Py-ya-afi.
80)87.

4.

Prove the identity

(a

+ + 7) (a + (aft + or 7) (a

2
-f

j8

+ vy) =

a3

+ )8 3 + 7 3 -

5.

Prove the identity

(o+

)3

a,

7)

(a

+ *>y)*=

(2a

- )8 - 7) (2j3 - 7 - a) (2y - a - 0).

Apply Ex.

2.

The Cube Roots of Unify.

6.

4:>

Prove the identity

Apply Ex.
7.

2,

and substitute for

o>

ur

its

value -v/-3-

Prove the identity

a
3

'

+ y 3 - 3a'V = (a3 +
j8'

+ 7 3 - 3 ajS-y) 2
2 7' = 7 + 2a.

where
a'
8.

= a2 + 207,

+ 27a,

Form

the equation whose roots are

m + n,
9.

urn

+ orw,

w-m +

tan.
(

Ans. x3 - Zmnx -

+ n 3 = 0.
)

Form

the equation whose roots are

l

+ m+n,
-4ns.

+ wi +

w,

#i

z 3 - 3fo2 + 3 (P - ftin) a; -

+ wH. 3 m3 + (i +

- Zlmn) = 0.

th important to observe that corresponding to the n n th roots of unity there are n n roots of any quantity. The roots

It

is

of the equation
*?

-a=

are the n n

tn

roots of a.

The

three cube roots, for example, of a are 3 /~ o 3 / 3 / a, w\/a, or /#,

where v^^ represents the

real cube root according to the ordinary Each of these values satisfies the arithmetical interpretation. It is to be observed that the three cubic equation x z - a = 0.

cube roots

may

be obtained by multiplying any one of the three

1, o>, or.

above written by

In

addition, therefore, to the real cube root there are

two

imaginary cube roots obtained by multiplying the real cube root by the imaginary cube roots of unity. Thus, besides the
ordinary cube root cube roots
3,

the

number 27 has the two imaginary

-

33
by

as the student can easily verify

10.

actual cubing.

Form

a rational equation which shall have

3

v/<3 + VQ2 + P* +
for a root
;

v/ Q.

where w
8.

=1

CompareEx.

46
11.

Roots and
Form an

Coefficients

of Equations.

equation with rational coefficients which shall have

for

root,

where

0i

1,

and

1.

Cubing both sides of the equation

and substituting x for

its

value on the right-hand

side,

we

get

z*-P-Q=Z
Cubing again, we have

Since Q\ and 6%
this equation are

may

each have any one of the values

1,

co, o>

2
,

the nine roots of

also that, since Q\ and 2 have disappeared from the final equation, it is which of these nine roots is assumed equal to x in the first instance. The resulting equation is that which would have been obtained by multiplying together the nine factors of the form x ^/P^/Q, obtained from the nine roots above

"We see

indifferent

written.
12. Form separately the three cubic equations whose roots are the groups in three (written in vertical columns in Ex. 11) of the roots of the equation of the pre-

ceding example.

"We can write these down from Ex.

then equal to w%/P~a>%/~Q
;

8,

taking

first

m and n equal to %/P, %

a;

and

finally equal to (S~%/P,

Ans.

Q x-P-Q = 0,

27.

Symmetric Functions of the Roots.

Symmetric

functions of the roots of an equation are those functions in which all the roots are alike involved, so that the expression is unaltered in value when any two of the roots are interchanged.

For example, the functions

of the roots, (the sum, the

sum

of the

products in pairs, &c.) with which we were concerned in Art. 23 are of this nature ; for, as the student will readily perceive, if
in

any

of these expressions the root

ai, let

us say, be written in

Symmetric Functions of

the Roots.

47
,

every place where a 2 occurs, and a- in every place where occurs, the value of the expression will be unchanged.

The

functions discussed in Art. 23 are the simplest symfirst

metric functions of the roots, each root entering in the

degree only in

any term of any one of them.

"We can, without knowing the values of the roots separately by means of the equations (2) 23 the values in terms of the coefficients of an infinite of Art.
in terms of the coefficients, obtain

variety of symmetric functions of the roots.

in a subsequent Chapter,

It will be

shown

resumed, that the roots can be so expressed. The examples appended to this most of which have reference to the simple cases of Article,
trate the usual elementary

the discussion of this subject is any rational symmetric function whatever of

when

the cubic and biquadratic, are sufficient for the present to illusmethods of obtaining such expresIt

sions in terms of the coefficients.

is

letter

usual to represent a symmetric function by the Greek attached to one term of it, from which the entire ex-

may be written down. Thus, if a, /3, y be the roots of a cubic, Sa 2 |3 2 represents the symmetric function
pression
2

/3

y + 0Y
7 +
2

products in pairs are taken, and each term 2 separately squared. Again, in the same case, 2a j3 represents

where

all possible

/3

+ a2 7 +

/3

/3

a + y~a + y

/3,

where

taken, and the

As

permutations of the roots two by two are root in each term then squared. 2 2 an illustration in the case of a biquadratic we take 2a /3 ,
all possible
first is
2

whose expanded form

o
a
s 2

as follows
2

/3

+ a 7 + a

j3Y +

j3

+ 72

2
.

By the
among

aid of the various symmetric functions which occur the following examples the student will acquire a facility

in writing out in all similar cases the entire expression the typical term is given.

when

48

Roots and

Coefficients

of Equations.

EXAMPLES.
1.

Find the value of 2cr

of the roots of the cubic equation

x3 + pzz
Multiplying together the equations

qx

0.

7 + 70 +

aft

q,

we

obtain

2o

-f

a&y
pq.

pq

hence
2.

2a 2
Find
for the

= 3r

same cubic the value of

a2

+ 72

Ans. 2 a 2

= p* -

1q.

3.

Find for the same cubic the value of

a3

&+
,

3
-

Multiplying the values of

2a and 2a2 we obtain

)8

a3
hence,

by Ex.

1,

+ T3 + 2a2 )8 = - p* + 2o3 = - p* + Zpq - 3r.

3

Ipq

4.

Find for the same cubic the value of

0V+7
"We easily obtain

a2

a2

2
.

ft

7)

from which
5.

2o

2
/3

2pr.

Find for the same cubic the value of

This

is

equal to

Ans. r
6.

- pq.

Find the value of the symmetric function

S 2 a/3

of the roots of the biquadratic equation

Multiplying together

aj87

aj85

+ 078 + ^75 = 4

r,

we

obtain

2 a- fiy +

0^78 = pr
4 s.

hence

2 a- fiy = pr

Excmiples.
7.

49

Find for

tlie

same biquadratic the value of the symmetric function

Squaring 2o,

we

easily obtain

2a2 =.p2 -2?.

8.

Find for the same biquadratic the value of the symmetric function

Squaring the equation

we

obtain

hence,

by Ex.

6,

Find for the same biquadratic the value of 2o3 0. this symmetric function, we take the two permutations ajS and /3a of 3 and 3 o of 2. We have similarly two the letters a, )8 these give two terms a
9.

To form

terms from every other pair of the letters tion consists of 12 terms in all.

a,

0, 7, 5

so that the symmetric func-

Multiply together the two equations

2<zj8

2',

2a2 =.p 2 -2;

and observe that

2a2 2aj8 = 2a3 j8 + 2a2 &y.

that results of the kind expressed by this last [It is convenient to remark here, equation can be verified by the consideration that the number of terms in both members of the equation must be the same. Thus, in the present instance, since

2a2

are in fact the 12 terms

contains 4 terms, and 2aj8 6 terms, their product must contain 24 and these which form 2o3 )8, together with the 12 which form 2a2 j8-y.]
;

Using the results of previous examples,

we

have, therefore,

10.

Find

for the

same biquadratic the value of

2 Squaring 2a and employing results already obtained,

,

2a 4 =j) 4 - 4p2 ? + 2 <? + Ipr - 4s.

11.

Find the value, in terms of the

coefficients, of the

sum

of the squares of the

roots of the equation

Squaring 2oi,

we

easily find
_

Pl

=2
.

hence

2ar=pi 8 -2j? 2
12.

Find the value, in terms of the

coefficient?, of the

sum

of the reciprocals of

the roots of the equation in the preceding example.

i;

50
From

Roots and
the second
02 03
last,

Coefficients

of Equations.
we have
n~ l
(

and

last of

the equations of Art. 23,

....

o/i

01 03

.... o

+.... +

01 02
1

....

o,,-i

l)
l)

pn-l>
J

01203

....

0=

n n p

dividing the former

by

the latter,

we have

01

tt2

03
1

On

Pn

s
01

= ~Pn-l pn

In a similar manner the sum of the products in pairs, in threes, &c. of the reciprocals of the roots can he found by dividing the 3rd last, or 4th last, &c. coefficient

by the

last.

13.

Find for the cubic equation

the value, in terms of the coefficients, of the following symmetric function of the
roots a, 0,

N. B. It will often be found convenient to write, as in the present example, an equation with binomial coefficients, that is, numerical coefficients the same as those which occur in the expansion by the binomial theorem, in addition to the literal Here the equation being of the third degree, the successive coefficients , i, &c.
numerical coefficients are those which occur in the expansion to the third power,
viz. 1, 3, 3, 1.

"We

easily obtain

14.

Express in terms of the coefficients of the cubic in the preceding example the

successive coefficients of the quadratic

where

a, 0,

are the roots of the cubic.

Here, in addition to the symmetric function of the preceding example,

to calculate also the

we have

two following

Ans. (o ("2
15.

2
i )

#2 .+ ( a o #3

a \ ^2) x +

(a\ as

a} = 0.

Find for the cubic of example 13 the value in terms of the

coefficients of

Since

2o-

-7= 3o- (o + + y) = 3o +

Examples.
the required value
is

easily obtained

by

substituting

<*o

for

x in the identity

a (x

-a

(*

-0) (x-y).

Ans. a<?(2a-l3
16. Find, in terms of the coefficients of the biquadratic equation

003*4 401Z3

+ 6a2 z2 4 4

3 o;

= 0,
:

the value of the following symmetric function of the roots

- 7) 2 (IB
Here the equation

~ 5) 2 4

(7

2
)

- 5)2 + (0

- 0)2 (7 (a

5)2.

written with numerical coefficients corresponding to the expansion of the binomial to the 4th power. The symmetric function in question
is is easily

seen to be identical with

Employing the

results of

examples 6 and

8,

we

find

17.

Taking the

six products in pairs of the four roots of the equation of

Ex. 16,
roots,

-and adding each product, e.g. o0, to that

which contains the remaining two

75,

we have

the three sums in pairs

7+
it is

a5,

70 4

05,

aj8

+ 75

required to find the values in terms of the coefficients of the two following
t

symmetric functions of the roots

(70 4 05) (o0

75)

(o0

78) (07

+ aS) 4

(07 + 08) (70

.

4 05),

(07

08) (70

4 05) (a0 4 75)

The former

of these

is

the

sum

of the products in pairs,

and the

latter the con-

tinued product, of the three expressions above given. As these three functions of the roots are important in the theory of the biquadratic, we [shall represent them

uniformly by the letters

of the coefficients for

A.,

/j.,

v.

"We have, therefore, to

find expressions in terms

The former

is

pv 4 v\ 4 AM> and */*" 2o2 07, and is easily expressed as follows

(cf.

Ex.

6)

The

latter

is,

when

multiplied out, equal to

and we obtain

after easy calculations the following

2

E2

52

Roots and

Coefficients

of Equations.
Ex.
16, the value of

18. Find, in terms of the coefficients of the biquadratic of

the following symmetric function of the roots

This

is also

To prevent any ambiguity

an important symmetric function in the theory of the biquadratic. in -writing this, or corresponding functions in which the

differences of the roots of the biquadratic enter,

we
ft,

explain the notation which will

be uniformly employed in this work.

Taking in
7,

circular order the three roots a,

7,

we have

the three differences

a,

j8

and subtracting
8,
ft

other differences

8,

8.

from each root in turn, we have the three We combine these in pairs as follows:

(0-7) (a-

3),

(7-o)OB-8),
is

(a-0)(7-S).

The symmetric function in question three taken as usual in circular order.

Employing the values
of A,
/*,

the product of the differences of these

in the preceding example,

we have

"We have, therefore, to find the value of

(3A

SctjS) (3/x

in terms of the coefficients of the biquadratic.

Multiplying this out, substituting the value of

of Ex. 17,
#o
3

2a, and attending

:

to the results

we - v)

obtain the required expression as follows

(2,u

(2A- n

- v- A) (2/ - A. - fi) = - 432 {a

a z a^+2ai

3-os2 -

2
i

4-23 }

The function

of the coefficients here arrived at, as well as those before obtained

in Examples 13, 15, and 16, will be found to be of great importance in the theory of the cubic and biquadratic equations.
19. Find, in terms of the coefficients of the biquadratic of Ex. 16, the value of the symmetric function

This

may

be represented briefly by 2

(a

2.

Ans. C 2 2(a-j8) 2
20.

Prove the following relation between the roots and

:

coefficients of the biqua-

dratic of Ex. 16

Theorems relating
28.

to

Symmetric Functions.

53

relating to Symmetric Functions. two theorems, with which we close for the present following the discussion of this subject, will be found useful in many in-

Theorems

The

stances in verifying the results of the calculation of symmetric functions.

(1)
.

The sum of the exponents of all

is

the roots in
to the

any term of any

suffixes in
coefficients.

symmetric function of the roots

equal each term of the corresponding value in terms of the

of,

sum of the
same

The sum here spoken

which

is

of course the

term of the symmetric function, and which

may

for every be called the

degree in all the roots of that function, will be subsequently defined (see Ch. XII.) as the weight of the symmetric function. The truth of the theorem will be observed in the particular cases

of the examples 13, 15, 16, 17, &c. of the last Article ; and that it must be true in general appears from the equations (2) of Art. 23, for the suffix of each coefficient in those equations is

equal to the degree in the roots of the corresponding function of the roots ; hence in any product of any powers of the coefficients
the

sum

of the suffixes

must be equal

to the degree in all the

roots of the corresponding function of the roots. (2). When an equation is written ivith binomial coefficients, the
expression in terms of the coefficients for any symmetric function of the roots, which is a function of their differences only, is such that

sum of the numerical factors of all the terms in it is The truth of this proposition appears by supposequal to zero. all the &e. to become equal to unity in coefficients ing i, ^2,
the algebraic
,

the general equation written with binomial coefficients,

a*xn
1
.

viz.,

-2

+ aH =

0.

&

The equation then becomes (x + l) n = 0, /. c. all the roots become equal hence any function of the differences of the roots
;

must
braic

in that case vanish,

and therefore
it
;

also the function of the

coefficients

which

is

equal to

but

this consists of the alge-

sum

of the numerical factors

when

in

it all

the coefficients

are made equal to unity. o, ffi, #2, &o. InjExs. 13, 15, 16, 18, 20 of Art. 27 we have instances of this theorem.

54

Roots and

Coefficients

of Equations.

EXAMPLES.
1.

Find in terms of p,

g,

r the value of the symmetric function

'

fly

ya

aft

where

a,

are the roots of the cubic equation

xs

+px2 + qx + r = 0.
<nq

Ans.
r
2.

-3.

Find

for the

same equation the value of

Ans. 24 r-p*.
3.

Calculate the value of

2

2a

3
;8

of the roots of the

;

same equation.
Ans. q*-

Here 2oj82a2

= 2a3

+ afiy2a*&

hence &c.

4.

Find for the same equation the value of the symmetric function

* is

easily obtained

3 by squaring 2o

(see

Ex.

3,

Art

27).

Ans.
5.

2p*

I2p*q

+ I1p

zr

+ 18^ 2 ^ 2 - ISpqr-

Find for the same equation the value of

/3+7

7+

a-|-j8

r-pq
6.

Find

for the

same equation the value of

+y

7+

7.

Find for the same equation the value of

2 fry
ft

-a2 + y-a

27a-j8 y + a-ft

2 aj8-72
a

+ 0-y' 3

8r

Examples.
8.

55
f

Find the value of the symmetric function 2

^^)

\a+

for the

same cubic

equation

Ans

(r-pqf
9.

Calculate in terms of p, q,

r, a

the value of

^ for the equation

x*

+px 3 + gxz + rx + s = Q.
;

Here

= 2| + 2 ^ 2a2a
2

and

7-

10.

Find the value of 2

of the roots of the equation

Ans
11.

Find

for the biquadratic of

Ex. 9 the value of

Compare Ex.
12.

22, Art. 24.

Ans. r z - p z s.
3
i)

Find the value of

2(oa +

(&-y)~ in terms of the

coefficients of the

cubic equation

= 0.
18

Ans.

^(ao2-ffi
o

2 2
)
.

13.

Find the value of the symmetric function 2

0102

of the roots of the

equation

The given function may be written

in the form

Ol

-7'O2 =+=+O;i
,

+ 02

(11 +
Ol

O2

On ;

[-

+ an
1

(11 +
(01

1 +.... + -

-1,

02

o,)

or2ai2
ai

hence &c.

Ans.-

fl\Pn-\
.

Pn

14.

Clear of radicals the equation

56

Roots and

Coefficients

of Equations.
t

and express the coefficients of the resulting equation in of the cubic of Ex. 1.
Ans.
15. If a,
/3,

in terms of the coefficients

3e--2*-2
Ex.
9,

7, 5 be the roots of the biquadratic of

prove

Substitute in turn each of the roots of the equation x 2

Art. 16, and multiply.

+1=

in the identity of

Prove the following relation between the roots and th equation of the n degree
16.
:

coefficients of the general

(ai

+ 1)

(a 2

2
)
. .
.

(a*

!)

(!

-pz + p*. -

2
.
.

.)

(P\

17.

Find the numerical value of

where

a,

)8,

7, 5 are the roots of the equation

- 5^ + 10=0.
Substitute in turn for x each root of the equation z?
18. If a, 0, 7, 5

+ 2 = 0, and

multiply.

Am.
be the roots of the equation

166.

prove
3
tfo

()8

+ 7)

(7 +

a)(a+)8)(a+5)

(0

+ 5)
is

(7

5)

= 16
(/u.

The symmetric function

Ayuy,

in question

equal to

+ v)

(v

+ \)
)8)

(A

+ ;u)>

or

2A

2fu/

where

A,

/t,

have the values of Ex.

17, Art. 27.

4

19. Calculate the value of the

symmetric function 2

(a

of the roots of the

biquadratic equation of Ex.

9.

Ans.
20.

3p*-16pz q + 2Qqz +4pr-16s.

be ex-

Show

that

when

the biquadratic

is

written with binomial coefficients, as in

Ex.

symmetric function of the preceding example pressed in the following form

:

18, the value of the

may

2 (a -

4
)

= 16 {48 (OQ a 2 -

2 2
i )

2 o

(o4 - 4 0i

+3

2 3 )}

21.

The
(a,

distances on a right line of

0)

the roots

and

(a,

')

of the

two pairs of points from a two quadratic equations

fixed origin are

+ c' =

prove that when one pair of the points are the harmonic conjugates of the other pair, the following relation exists
:

ac

+a'cA,

2W = 0.
,

22.

The

distances of three points

C on

a right line from a fixed origin

on the

line are the roots of the equation

Examples.
find the condition that one of the points ,4, B,

57
bisect the distance

C should

between

the other two.

Compare Ex.
four points 0,

15, Art. 27.

Ana. d*d- 3abc+ 243 =

0.

23. Retaining the notation of the preceding question, find the condition that the

A, B, C should form a harmonic

division.

Ans. ad 2 -3bcd + 2c3 = Q.

This can be derived from the result of Ex. 22 by changing the roots into their can be easily calculated independently.
24. If the roots
(o, ft,

reciprocals, or it

7, 5) of the equation

are so related that o

5,

8,

7-8
+
2 bed

are in harmonic progression, prove the relation

among

the coefficients
ace

- ad* -

b- e

- c3 =

Compare Ex.
25.

18, Art. 27.

Form

the equation whose roots are

fly

+ wya + o> 2 aft

7 + (o ya + waft
2

where

a>

1,

and

a,

ft,

are the roots of the cubic

ax?

3 bx~

Ans.

+ 3 ex + d = 0. - b z z~ + (ad- be] x + (bd-c~)=0. (ac

)

Compare Exs. 13 and

26. Express
(2

14, Art. 27.

7 - 70

aft)

(2ya -aft- fty)

(2

aft-fiy- ya)

as the

sum

of

two cubes.
Ans.
(fty

+ wya + rfaft) 3 -f (fty + &> 2

Compare Ex.
27. Express

5,

Art 26. x+y

+ 2 )3 +
xyz,

x + uy -f ors) 3 43

(x

+ o>2 y + wz) 3
-4*

z in terms of x3 + y

+ z3 and

where

= 1.

28. If

find

i, Y, Zin terms of x, Apply Example 4, Art.

y, z; x', y',

z'.

26.

Z=xz' + yx'
29. Resolve
(a

+ ft +

7)

0)87

(/3y

+7
ft,

into three factors, each of the second degree in a,

7.
2

Ans.

(a

- fty)

(3

ya)

2 7 -

Compare Ex.

18, Art. 24.

58

Roots and

Coefficients

of Equations.
:

30. Resolve into simple factors each, of the following expressions

(1).

(j8-

(2).

(3_
An*.
(1). (2).

(2a--7)(2-7-a)(27-a-0).
-908- 7 )( 7 -)(-jB).

31.

Find the condition that the cuhic equation

x3

-px z + qx-r =
a

should have a pair of roots of the form a

the roots in that case.
If the real root
roots,
is b,

^/-

and show

how

to determine

we

easily find,

by forming the sum

is

of the squares of the

pz

2q =

b2 .

The

required condition

32. Solve the equation

x3

-7*2

4-

20*-24 =
31.
Aiis.

0,

whose

roots are of the

form indicated in Ex.

Boots

3,

and 2 +

33.

Find the conditions that the biquadratic equation

a;

px* + qx
a

rx + s

=
1.

should have roots of the form a

conditions

\/-

1,

/-

Here there must be two

among

the coefficients, as there are only two independent quantities

involved in the roots.

Am.
34. Solve the biquadratic

2
^?

-2^ = 0; r*-2q$ = Q.

z4

+ 4s 3 +8z2 - 120* + 900 = 0,

33.

whose

roots are of the

form in Ex.

Ans.
35.
/ Ifa+#V - 1 k ea root

+ 3 /^T, -

of the equation

prove that 2a will be a root of the equation

#3 + qx
36.

r = 0.

Find the condition that the cubic equation

z 3 +pxz + qx + r =

should have two roots

o,

)8

connected by the relation

ojS

+ 1 = 0.
l

Ans.

Examples.
37.

59

Find the condition that the biquadratic

should have two roots connected hy the relation

oj3

+ 1 = 0.
+ s*= 0.

The

condition arranged according to powers of

* is

+pr +
38.

r*

+ (p* +pr - 2q - 1) s + (q - a2
2
)

1) s*

Find the value of 2

(ai

0304

a n of the roots of the equation

This

is

readily reducible to Ex. 13.

Ans.
39. If the roots of the equation
.

(-

aoz* + na\ z"- 1

'
t

-f

a^x"-*

+.... + a n =

be in arithmetical progression, show that they can be obtained from the expression

0~
by giving
0, 2, 4,

to r all the values 1, 3, 5, ....

n-

1,

when n

is

even

and

all

the values

6 .... n

- 1, when n

is

odd.

40. Representing the differences of three quantities a, 0,

7 by

01,

lt

71, as

follows

01

= )8 - 7,

= 7 - a,

71

= a - j8

prove the relations

ox

Pi 5

7i

= I {ai 2 + )8i 2 + 7i 2 }aii7i.

01, 0i,

These results can be derived by taking

71 to be roots of the equation

-r=
(where the second term
is
3
,

absent since the

4
,

sum

of the roots

= 0), and calculating the

symmetric functions 2ai 5ai tended to form 2ai 6 2oi 7 &c.

,

2ai 5 in terms of q and r. The process can be exThe sums of the successive powers are, therefore,

capable of being expressed in terms of the product 01 #171 and the sum of squares 2 2 2 ai + )8i + 7i ; the former being equal to r, and the latter to - 2 (/Ji 71 + 71 01 -f ai 0i), 3 or - 2q. These sums can be calculated readily as follows By means of x = r gx,
all
:

and the equations derived from this by squaring, cubing, &c., and multipljing by x or a;2 any power of x, say XP, can be brought by successive reductions to the form + Bx + Cx~, where A, B, C are functions of q and r. Substituting ai, /3i, 71, and
,

adding,
this

we

way

find 2ai?

3^4

2?^.
(?

The
3

student can take as an exercise to prove in

= 7?V,

2ai

n= llqr

-r2 ).

CHAPTER

IV.

TRANSFORMATION OF EQUATIONS.
29.

Transformation of Equations.
knowing the values
it

We

can in

many

instances, without

of the roots of

an equa-

tion in terms of the coefficients, transform

stitutions, or

hy elementary sub-

by

the aid of "the symmetric functions of the roots,

into another equation whose roots shall have certain assigned transformation of this relations to the roots of the proposed.

nature often facilitates the discussion of the equation. proceed to explain the most important elementary transformations of equations. 30. Roots with

We

tion into another

Signs changed. To transform an equawhose roots shall be equal to the roots of the
let ai, a 2 , a 3 ,
.

given equation with contrary signs, roots of the equation

a n be the

Xn + p,Xn

~l

+ p 2 Xn

~Z

pn-iX

+p n =
-

0.

We have then the identity

xn

+p

xn

~l

+p

xn

~2

+pn_i% +p n = (xpn-iy +p n =
.

ai) (x

a 2)

(x

a n)

changing x into
y -piy
n
n

y,

we
.
.

have, whether n be even or odd,

(y

^+P*y

n ~z

i)

(y

2)

(y

+ a n ).

The polynomial
whose roots are transformation

ai,

in y equated to zero is, therefore, an equation - a2 - a n and to effect the required

, . . ;

we have only

to change the signs of every alternate

term of the given equation beginning with the second.

EXAMPLES.
1
.

Find the equation whose

roots are the roots of

#6 + 7^4

7^3

_8

+ +1=
a;

with their signs changed.

Ans. #5 -7a?*+

7^+

8z 2 + x

= 0.

To Multiply Roots by a Given Quantity.

2.

61

Change the signs of the roots of the equation

= 0.
[Supply the missing terms with zero
coefficients.]

Am.
31.

xT

To

Multiply the Roots .by a Given Quantity.

equation whose roots are
ai,

To transform an

a2

into anointo

ther whose roots are

ma^

ma*,

ma n we change x
,

m
,

in

the identity of the preceding Article.

Multiplying by

we

have
y
n
-f

Mpiy"- +

mp
z

n ~z

rn^p^y + mnpn
(y

= (ymai)

- ma z ---)

(y

- ma n ).

Hence, to multiply the roots of an equation by a given quantity m, we have only to multiply the successive coefficients, beginning
with the second, by m,

z
,

m?,

mn

present transformation is useful for the purpose of rethe coefficient of the first term of an equation when it moving is not unity and generally for removing fractional coefficients
;

The

from an equation. .If there is a coefficient c of the first term, the we form the equation whose roots are a Q ai, a a 2 oa n transformed equation will be divisible by a and after such divitf
, .
.

sion the coefficient of xn will be unity.

When there are fractional

by multiplying
the roots
all

coefficients,

we can

common

multiple of

by a quantity the denominators of the fractions.

get rid of them which is the least

In
least

many common

cases multiplication

by

a quantity

less

than the

multiple will be sufficient for this purpose, as will in the following examples appear
:

EXAMPLES.
1
.

Change the equation

3*4

4*3

+ 4^2 _ 2 * +

=
-

into another the coefficient of

whose highest term

3.

will he unity.
x*>

We
2.

multiply the roots by

Am.

4^ +

12.i

- 18* +

27

= 0.

Remove

the fractional coefficients from the equation

^_^ + ?,_l =
Multiply the roots by
6.

0.

Ant.

a?

- 3z2 + 24* - 216 = 0.

62
3.

Transformation of Equations.
Remove
the fractional coefficients from the equation

By noting the factors which occur in the denominators of these fractions, we observe that a number much smaller than the least common multiple will suffice to
remove the
fractions.
:

If the required multiplier

is

m,

we

write the transformed

equation thus

it is

evident that

if

m he taken =

6,

each coefficient will become integral

6.

hence

we

have only to multiply the roots by

An*, *3 - 15*2 -14* + 2

= 0.

4.

Remove

the fractional coefficients from the equation

The student must be

terms with zero

careful in examples of this kind to supply the missing

coefficients.

The

required multiplier

is

10.

Ans. ** + 30*2
5.

+ 520* + 770 = 0.

Remove

the fractional coefficients from the equation

Ans. * 4

- 25*3 + 375*2 - 11700 = 0.

32. Reciprocal Roots and Reciprocal Equations. To transform an equation into one whose roots are the reciprocals
of the roots of the proposed equation,

we change x

into

- in the

identity of Art. 30.

reductions,

This substitution gives, after certain easy

or

Pn
hence,
if

a2

in the given equation

we

replace x

by

-,

and multiply

the resulting polynomial in y equated to zero will have for roots the reciprocals of ai, a*, . . . . a*.

by y

Reciprocal Equations.

63

There

is

a certain class of equations which remain unaltered

reciprocal

when x is changed into its reciprocal. These are called The conditions which must obtain among equations.
ficients of

the coefclass

are,

an equation in order that it should he one of this by what has been just proved, plainly the following
:

Pn-\

Pn-z

^.-* **.-*,-*.
P
* pn =

PI

The
equal to
(1).

last of these conditions gives

1,

OT p n

l.

Reciis

procal equations are divided into two classes, according as p n

1, or to

1.

In the

first

case

we have
Pn-z

the relations
.
.

Pn-i

= Pl,

=P

Z,

Pi =Pn-i

which give

rise to

the first

class

of reciprocal equations, in ichich

the coefficients of the corresponding terms taken from the beginning and end are equal in magnitude and have the same signs.
(2).

In the second
Pn-l

case,

when^?, =
t

1,

we have

= -Pi

giving rise to the second

class of reciprocal equations, in which corterms responding counting from the beginning and end are equal in but It is to be observed that in this magnitude different in sign.

the degree of the equation is even, say n = 2m, one of the conditions becomes p m = -p m9 or p m = ; so that in reciprocal
case

when

equations of the second

class,

whose degree

is

even, the middle

term

is

absent.

If a be a root of a reciprocal equation,

- must
a

also be a root,

for it is a root of the transformed equation,

and the transformed

hence the roots of a
;

equation

is

identical with the proposed

reciprocal equation occur in pairs, a,

/3, 7*

&o-

When

the
;

degree

is

and

it is

in fact obvious

odd there must be a root which is its own reciprocal from the form of the equation that - 1

then a root, according as the equation is of the first or second of the above classes. In either case we can divide off by
is

or + 1

64
the

Transformation of Equations.

known

factor (x

+ 1 or x -

1),

and what

is left is

a reciprocal

equation of even degree and of the first class. In equations of the second class of even degree x 2 - 1 is a factor, since the equation may be written in the form

2 1, this also is reducible to a reciprocal dividing by x of even degree. the first class Hence all reciprocal of equation

By

equations
even,

may be reduced to those of the first class whose degree is and this may consequently be regarded as the standard form

of reciprocal equations.
EXAMPLES.
1
.

Find the equation whose

roots are the reciprocals of the roots of

Ans.
2.

2/-5y 3
and of
first class

Reduce to a

reciprocal equation of even degree

Ans. z* + -x*

-636

xz + -

33.

To Increase
To

or Diminish the Roots by a Given

transformation

Quantity.

effect this

we change

the vari-

able in the polynomial f(x) by the substitution x = y + h; the resulting equation in y will have roots each less or greater by //

than the given equation in

tive.

x,

The

resulting equation

according as h is (see Art. 6)

is

positive or nega-

There

is

mode
is

of formation of this

practical purposes

much more

equation which for convenient than the direct cal-

culation of the derived functions,

of the given quantity h.

This we proceed to explain.

and the substitution in them Let the

proposed equation be

flfc**"

+ fl_i# + a =

To Increase

or Diminish the Roots.

to be

65

and suppose the transformed polynomial in y

A
since

y + A,y

n~ l

^+

A^y + A n

y = x-h,

this is equivalent to
n

(x

h}

(v- A)- +

A^(x -h) + A n

which must be identical with the given polynomial. We conclude that if the given polynomial be divided fyy x - h, the remainder
is

ny

and the quotient

if this

again be divided by x

h,

the remainder

is -4n_i,

and the

quotient

Proceeding in this way, we are able by a repetition of arithmetical operations, of the kind explained in Art. 8, to calculate in succession the several coefficients n An-\<> &c., of the trans-

formed equation

the

last,

being equal to a Q

in a subsequent Chapter that the best practical

method

It will appear of solv-

ing numerical equations is only an extension of the process employed in the following examples.

EXAMPLES.
1.

Find the equation whose

roots are the roots of

each diminished by 4.

The

calculation is best exhibited as follows

-5
4
1

-17
12
-

11

4
3 12

20

4
3

60

15

55

28

~~
7

48

4 11

66
x

Transformation of Equations.
Here the first division of the given polynomial by x and the quotient x 3 - xz + Zx - 5 (cf. Art. 8). we 4, get the remainder 55 (= -^3), and the quotient
;

4 gives the remainder

Dividing this again hy x* + 3x + 15. Dividing
this

9 (= At),

we get the remainder 43 (= Az), and quotient x + 7 and dividing Ai=ll, and AQ = 1 hence the required transformed equation is
again,
;

we get

Find the equation whose

roots are the roots of

each diminished by

3.

Removal of Terms.
calculation

67

may be

exhibited in two stages, as follows, the broken lines marking the

:

conclusion of each stage

13

12

100

1740 1728

34560

87
100 187
100

34567
19122

3740

5468
906

53689

287
15

6374
951

302
15
~317
15

7325

332

Am.
34.

by*

+ 332y 2 + 7325 y+ 53689 = 0.

of the chief uses of the
is

Removal of Terms.

One

transformation of the preceding Article

to

remove a certain

Such a step often facilitates specified term from an equation. its solution. the transformed equation in descending Writing
powers of
y,

we have
/*, (

If h be such as to satisfy the equation na h + a = 0, the transformed equation will want the second term. If h be either of

the values which satisfy the equation

Mrc-l) 1.2

(n

1)

0,

the transformed equation will want the third term ; the removal of the fourth term will require the solution of a cubic for h and
;

so on.
0,

To remove
which
is

the last term

we must

solve the equation

the original equation

itself.

68

Transformation of Equations.

EXAMPLES.
1.

Transform the equation

into one

which

shall

want the second term.

ai

naoh + Diminish the roots by 2. 2. Transform the equation

into one

=Q

gives h

= 2.
Ans. y*
8y
15

= 0.

which

shall

want the second term.

Increase the roots

3.

by

2.

Ans. y*
a4

- 24 y2 + 65 y - 55 = 0.

Transform the equation

_ 4^3 _ i s X*- - 3# + 2 =
third term.

into one

which

shall

want the
is

The quadratic

for h

6A2

-12A-18 = 0,
3,

giving h =3, h

=-1.

Thus there

are

two ways of

effecting the transformation.

Diminishing the roots by

(1)

we

obtain
s

y+8y we
y*

Increasing the roots

by

1,

obtain

(2)

35.

Binomial
:

Coefficients.

In many algebraical pro-

cesses it is

found convenient to write the polynomial /(#) in the following form

..
.

in which each term

is affected,

in addition to the literal coef-

ficient, with the numerical coefficient of the corresponding term n The in the expansion of (x + l) by the binomial theorem.

student will find examples of equations written in this way on The form is one referring to Article 27, Examples 13 and 16.
to

which any given polynomial can be

at once reduced.
:

We now

adopt the following notation

1

Un = a xn + nciix*-

-\J.
.
<

a^-

+ nan-ix+ an

with the suffix n to represent the polynomial of thus using the nth degree written with binomial coefficients.

Binomial

Coefficients.
&c.,

69
1,

"We have, therefore, changing n into n 1

(n

1)

a^-' +

(n

1)

a^

= a Q x3 +

a x a
.

One advantage

of the binomial

form

is,

that the derived

functions can be immediately written down. function of n is, plainly,

The

first

derived

or n

Un

-i

so that the first derived function of a polynomial re-

presented in this way can be formed by applying to the suffix of 7 the rule given in Art. 6 with respect to the exponent of the
variable.

Thus, for example, the

first

derived of Z74

is

formed

by multiplying the function by 4, and diminishing the suffix by unity it is, therefore, 4 Z7"3 as the student can easily verify.
; ,

We

proceed

now

to prove that the substitution of

ij

+ h for

x transforms the polynomial

Vn

or

into

where
AU) A\y Afy
are the functions "which result
J7 ,
i.e.
ZTi,
.
.

An

-\i

An
for

by substituting h
CT-i,
t

x in

Kj

VH

=a

= aQ h + a

l,

Az = aoh

^-2aji + a2 t &c.
suffixes, as

Kepresenting the derived functions of f(h) by

70

Transformation of Equations.
6,

explained in Art.
mation, viz./(y +

we may

write the result of the transfor-

^), in

the following form:

f(h)

is

An

the result of substituting li for x in n; its first derived/! (A) is, by the above rule,
is

it is,

therefore,
;

nA n-i

the

first

derived of this again

substitutions,
to write

n(n-l) A n_ z and
;

so on.

Making

these

we have the result above stated, which enables us down without any calculation the transformed equation.

EXAMPLES.
nd the
result of substituting

y+ h

for

x in the polynomial

Ant.

The student

will find

it

calculation explained in Art. 33,

a useful exercise to verify this result by the method of which may often be employed with advantage in

the case of algebraical as well as numerical examples. 2. Remove the second term from the equation

We must diminish the roots by a quantity h obtained from the equation

Substituting this value of h in

3

z,

and

3,

the resulting equation in y

is

_
o

3. Find the condition that the second and third terms of the equation Vn = should be capable of being removed by the same substitution. Here A\ and At must vanish for the same value of h and eliminating h between them we find the required condition.
;

Ans.
4. Solve the equation

aottz

2
i

= 0.

by removing its second term. The third term is removed by the same

substitution,

which gives

The
The
required roots are obtained

Cubic.

71
root of the latter

by subtracting 2 from each

equation.
5. Find the condition that the second and fourth terms of the equation Un = should be capable of being removed by the same transformation. Here the coefficients A\ and AS must vanish for the same value of A eliminat;

ing h between the equations

we

obtain the required condition

N.B.

"When

this condition holds

among

tion its solution is reducible to that of a quadratic

the coefficients of a biquadratic equafor when the second term is ;

;

removed the resulting equation of x can be obtained.

6.

is

a quadratic for y2

and from the values of y those

Solve the equation

*4 + 16 x3

+ 72z 2 + 64* - 129 =

by removing its second term. The equation in y is

y4_ 247/2-1=0.
7

Solve in the same manner the equation

& + 20s
8.

+ 143*2 + 430* + 462 = 0.

Ans. The
roots are

- 7, -3, -$*/Z.

Find the condition that the same transformation should remove the second
terms of the equation

and

fifth

Un = 0.
Ans.
ao
3

4<ro

i3+ 6aoi 3

3ai

= 0.

36.

The

Cubic,

On

account of their peculiar interest,

we

and the next following Articles the equations of the third and fourth degrees, in connexion with the transformation of the preceding Article. When y + h is subshall consider in this

substituted for x in the equation

Q,

(1)

we

obtain

A
where AI,

0,

A A
z,

have the values of Art. 35.

is

If in the transformed equation the second term

absent,

AI =

0,

or h =

rtj
.

72

Transformation of Equations.

Substituting this value for h in Art. 35,

and

z,

we

find, as in

Ex.

2,

- 3a

a qz + 2a^

hence the transformed cubic, wanting the second term,

o
8 y +

is

(a

a z - a?} y +

1
.

(a *a 3

- 3a

^ a* + 2aS) = 0.

The
mary to

functions of the coefficients here involved are of such

it is

importance in the theory of algebraic equations, that

represent the notation

custo-

them by

single letters.

We accordingly adopt

a az - a*

and write the transformed equation in the form

If the roots of this equation be multiplied

by a

it

becomes
(3)

G=Q:
cussion of the cubic.
to aQ y or a Q x + a v
identical with
3 (o* + a,) + 3ff (ax + a,) + G\= 0,
.

a form which will be found convenient in the subsequent

dis-

The variable, s, herein contained is equal The original cubic multiplied by a z is in fact

as the student can easily verify. If the roots of the original equation be a,

j3,

7, those of the

transformed equation (2) will be

_

-,
flo

+ -,
flo

-; 7+a

or,

snce

they

may

be written as follows

-/3- 7 ),

$ (2/3-y-B),

The Biquadratic.

;.;

"We can write down immediately by the aid of the transformed equation the values of the symmetric functions

S(2a-0- 7 )(20-y-a),

(2a-/3-y)(2/3- 7 -a)(2y-a-/3)

of the roots of the original cubic.

The

latter will

be found to

agree with the value already found in Ex. 15, Art. 27. may here make with regard to the general equation an important observation that any symmetric function of the roots

We

a, /3, y, 8,

&c., which

be expressed by the transformed equation wanting the second term.

obvious,
since the difference of
is

a function of their differences only, can the functions of the coefficients which occur in
is

This

is

any two roots

a',

|3'

of the

equal to the difference of the two correthe roots of a, )3 original equation and any symmetric sponding function of the roots a', /3', 7', S', &c., can be expressed in terms

transformed equation

of the coefficients of the transformed equation. For example, in the case of the cubic, all symmetric functions of the roots which contain the differences only can be expressed as functions of
fl
,

-ffj

and G.

Illustrations of this principle will be

'

found

among
37.

the examples of Art. 27.

The Biquadratic,
is

The transformed

equation, want-

ing the second term,

in this case

where
Article

A
;

and A 3 have the same values as in the preceding and where A is given by the equation 2 2 - 3flA # <M 4 = a 3 #4 - 40 a\ a* + 60 #2
i

The transformed equation

/?

is,

therefore,

/+ #0 Htf+
9

Gy+
GQ &0

>
H

2 tf4-4a 0i a, + 6

a 1 flf,-3a l

=0.

we pleased represent the absolute term of this and G, and have thus three funcequation by a symbol like
might
if

We

tions of the coefficients, in terms of

which

all

symmetric func-

tions of the differences of the roots of the biquadratic could be exthis term as pressed. It is more convenient, however, to regard

74

Transformation of Equations.

composed of Iff and another function of the coefficients determined in the following manner We have plainly the identity
:

3 4

- 40

0!

+ 60

0i

-3aS = a

(a

- 40!

+ 30 22 ) - 3 (0

2 2
i

This involves
viz.,

H, and another
4

function of the coefficients,

-40!

+ 30 22

which

is

of great importance in the theory of the biquadratic.

is

This function
3

represented by the letter J, giving

04

40

2
1

-t-

60

0i

- 30! 4 s

2
<2

3J? 2

The transformed equation may now be

written
r2

*+
We

^
6H
,

+**
;

J-3J2

-ir--2

can multiply the roots of this equation, as in the case of the cubic of Art. 36, by and obtain
s
4

+ 6J?s + 4Gz +

1- 3H

0.

(2)

This form will be found convenient in the treatment of the

same

algebraical solution of the biquadratic. as in the case of the cubic, viz., a x +

The
0i.

variable

is

the

The

original bi-

quadratic
(a Q x

is

in fact identical with

4

0!)

+ 6jff(0 # + 0!) 2 + 4(7 (a,x + a,) + a? I- 3H* =

3

0,

removed from this latter equation. function of the roots of the original biquaAny symmetric dratic which contains their differences only can therefore be
after the factor
is

If, G, and /. expressed by If the roots of the original equation be the transformed (1) will be, as is easily seen,
,

a, )3, 7, 8,

those of

The sum
=
00
r
;

of these =

the

sum

of their products in pairs

the

sum

of their products in threes =

- ;

and

for their

#0

HomograpMc
continued product

Transformation.

75

we have the equation

There
wish

is

another function of the coefficients to which

call attention, as it will

we

now to

be found to he of great imIt is

portance in the subsequent discussion of the biquadratic. the function arrived at in Ex. 18, Art. 27, viz.,
a a z aA + 2a l a z a 3 - a^a^ - a*aA a.z *.

This

is

shows that

denoted by the letter J. The example in question it is a function of the differences of the roots. It

must, therefore, be capable of being expressed in terms of a 0t jET, 6r, and /. have, in fact, the identity

We

which the student can

easily verify.
:

Or

this relation can be derived as follows

,

Whenever a

function of the coefficients a

a\,

# 2 &c. is the expression of a

>

function of the differences of the roots, it must be unaltered by the transformation which removes the second term of the equation,
2

hence

its

value

is

unaltered

when we change

a^

into zero,

into

2,

# 3 into
1

3,

&c.

Thus

#0^204

+ 2a a z a 3 - a^a* - a*a - a* = a^A^A^ - a

A^ - A*

their values in terms of H, G, I, we 2 , A^ substituting for obtain the above easily identity, which will usually be written in the form G* + 4H* - fl 2 (HI- a J).
38.

Homographic Transformation.
is is

The transforma-

tion considered in Art. 33 in which x

a particular case of the following, connected with the new variable y by the equation

\X +

fj.

If

X=

fi

=-

h, X'

0,

p.'

1,

we have y = x-h,

as in Art.

33

Solving for x in terms of

y,

we have

76

Transformation of Equations.
This value can be substituted for x in the given equation,

and the resulting equation of the nth degree in y obtained. Let a, /3, y, S, &c., be the roots of the original equation, and a', /3', y', S', &o., .the corresponding roots of the transformed
equation.

From

the equations

Xa + M

we

easily derive the relation

,

with corresponding relations for the differences of any other

pair of roots.
If

sponding

roots,

we take any four roots, and we obtain the equation

the four corre-

the roots of the proposed equation represent the distances of a number of points on a right line from a fixed origin

Thus,

if

on the line, the roots of the transformed equation will represent the distances of a corresponding system of points, so related to the former that the anharmonic ratio of any four of one system
is

the same as that of their four conjugates in the other system. It is in consequence of this property that the transformation is
It
is

called homographic.

sidered, in

important to observe that the transformation here conwhich the variables x and y are connected by a relation

of the form

Axy + Ea;+Cy + D =
is

0,

variable corresponds one,

39.

the most general transformation in which to one value of either and only one, value of the other.

Transformation by Symmetric Functions.

Sup-

required to transform an equation into another whose roots shall be givn rational functions of the roots of the pro-

pose

it is

Let the given function be (a, /3, 7 posed. involve all the roots, or any number of them.

.),

where

may
all

We

form

pos-

Transformation by Symmetric Functions.

sible

77

combinations

(aj3y),

(a/3S), &c., of the roots of this type,

and write down the transformed equation

as follows

will

"WTien this product is expanded, the successive coefficientsjof be symmetric functions of the roots a, ]3, y, &c., of the

given equation ; and may therefore be expressed in^terms of the coefficients of that equation.
EXAMPLES.
1.

The
7

roots of

are a,

find the equation whose roots are a 2 , transformed equation to he the Suppose
/3,
;

2
jS
,

z
.

then

-P=
,.

2 2 2 2 2 and we have to form the symmetric functions 2a Sa ^, o /3 7 of the given equaobtain "We easily tion.
,

the transformed equation

is,

therefore,

2.

Find in the same case the equation whose

Ans.
3

roots are
z

a3

3
,

/3

3
.

y + (p* 3pq + 3r) y + (q* Zpqr +

f

3r2) y

+ r 3 =,0.

3.

If a,

/3,

7, 5 be the roots of

xt+pxt + qyp + rz-l s=0;

find the equation

whose

roots are

a2,

2
j8
,

2
,

S2 .

Let the transformed equation be

y
then

+ .fy 3 + Qy* + Ry + S = 0,
,

-P=2a3
Compare Exs.
8,

Q = 2a 2 /32

Z = 2a 2
(q*

3
/3

2
,

5=a2 /3
(r

72 5

2
.

17, Art. 27.

^w*. y*
4. If a,
|3,

- (p- - 2q) y 3 +

- 2pr + 2s} y2 -

2qs) y

0.

7, S be the roots of

OQX*
find the equation

+ a\& + 6
ju,

*2
',

+ 4 3 ^+ a* =
+ 75.

whose

roots are A,
/37

viz., a/3

+ a5,

ya + (38,

See Ez. 17, Art. 27.

Ant. y 3 o

+
<*o

--7 (4i ^ 3 -a Q ai)y

(2fl

-3

2^4

-f

2a 1 'a 4 )

= 0.

78
5.

Transformation of Equations.
Show
that the transformed equation,

when the roots

of the resulting cubic of

is

Ex. 4 are multiplied by ^o, and the second term of the equation then removed,
Z 3_

40. Formation of the Equation whose Roots are any Powers of the Roots of the Proposed. The method of
effecting this transformation by symmetric functions, as exmuch plained in the preceding Article, is often laborious.

simpler process, involving multiplication only, can be employed. It depends on a knowledge of the solution of the binomial equaThis form of equation will be discussed in the tion xn - 1 = 0.

next Chapter.
to the student

The general

process will be^ sufficiently obvious

from the application to the equations of the 2nd and 3rd degrees which will be found among the following ex:

amples

EXAMPLES.
1.

Form

the equation whose roots are the squares of the roots of

X n + piX"- l +p2X n

-Z

+ .... + p n-l
the identity

To

effect this transformation,

we have

x+pix

n- l

+p z x- +.
x,

.+p^ix+p n = (x-ai

changing x into

we

derive, as in Art. 30,

xn-pixn -i + p
multiplying,

we have

it is

evident that the

;

first

member

of this identity contains,

when expanded, only

finally
2
).

even powers of x

we may then

2 replace x

by

y,

and obtain

-. ..

= (y-ai 2 )(y-a 22 )...(y-a

The

first

N.B.

the required transformed equation. This transformation will often enable us to determine a limit to the numof this equated to zero
is

member

her of real roots of the proposed equation. For, the square of a real root must be and therefore the original equation cannot have more real roots than the positive
;

transformed has positive roots.

2.

Find the equation whose

roots are the squares of the roots of

Am.
The
latter equation,

by

Descartes' rule of signs, cannot have

more than one

roots.

positive root; hence the former

must have a pair of imaginary

Examples.
3.

79

Find the equation whose roots axe the squares of the roots of the equation

Ans. y 5 + 2y 4
It follows

+ 5y 3 -f

2
3,y

2y

- 9 = 0.
must have

from Descartes' rule of signs that the

1

original equation

four imaginary roots. 4. Verify by the method of Ex.

6.

the Examples 1 and 3 of Art. 39.

Form

the equation whose roots are the cubes of the roots of

It will be observed that in

Ex.

/(#), the given polynomial,

and/(-

x)

the process consists in multiplying together the variables involved in these being those
:

which are obtained by multiplying x by the two roots of the equation # 2 1 = 0. In the variables inthe present case we must multiply together /( x), /(&>#), /(a> 2 )
:

volved being obtained by multiplying x by the roots of the equation x* transformation may be conveniently represented as follows
:

= 0. The

Write the polynomial/^) in the form

which we represent, for brevity, by

where P, Q, and R are "We have then

all

functions of x3 .

(x-a^)(x-az) ---- (x-a n ).

Changing, in this identity, x into (ax and at*x successively,

(1)

we

obtain
(2) (3)

P+<axQ-+a>~x~=(<*>x-ai)(utX-a2)

(wx-ctn),
2
.

P+ w

xQ + wx'tR s

(<D

ZX

- 01) (<*>*x - 02)

(o

#-

an),

3 since P, Q, and E, being functions of x , are unaltered. of (1), (2), (3), and attending to the results members both together Multiplying

of Art. 26,

we

obtain

The
only.

first

We

of this identity contains x in powers which are multiples of 3 3 can, therefore, substitute y for x and obtain the required transformed

member

equation.
6.

Find the equation whose

roots are the cubes of the roots of

Ans.
7. 8.

y*<

+ 14y 3 + 60y2 + 6y + 1 = 0.

39. Verify by the method of Ex. 6 the result of Ex. 2 of Art. Form the equation whose roots are the cubes of the roots of

Ant. a

80
41.

Transformation of Equations.

lem of whose roots are connected by a given

Transformation in General. transformation we have to form

In the general proba

new
<

equation in y,
(x,

relation

y)

with

the roots of the proposed equation f(x) = 0. The transformed equation will then be obtained by substituting in the given

equation the value of x in terms of y derived from the given = relation or, in other words, by eliminating x be(x, y) tween the two equations /(#) = 0, and (x, ij] = 0. For example,
;

suppose it were required to form the equation whose roots are the sums of every two of the roots (a, |3, 7) of the cubic

We have here

x3 -px* + qx - r =

0.

The equation (#, y) = is in this case y=p-x\ for when x takes the value a, y takes one of the proposed values and when x takes the values j3 and y, y takes the other proposed values.
;

The transformed equation is therefore obtained by p - y for x in the given equation.

EXAMPLES.
1.

substituting

If

o,

)8,

7 be the

roots of the cubic

form the equation whose

roots are

**+;
Here

and the given

substituting

relation is

xy =

+r
0.

the transformed equation

is

then obtained by

-- for x inf(x) =
y

Ans.
2.

rywhose
roots are

Form, for the same

o

cubic, the equation

a-y,

0)8

&y,

Py +

ay.
0.

Substitute

for x.

9-y

Ans. y* - 2qyz + (pr + j2 ) y + r2 - pqr =

Equation of Squared Differences of a Cubic.

3.

81

Form,

for the

same

cubic, the equation

5

whose

roots are

8 + y-a

7+a-

7
'

+ 0-7"

Substitute- for*
Ans. Cp 3
t

,4.

If a,

/3,

7 be the

roots of the cubic

prove that the equation in y whose roots are

7- a2 + 7 - 2a' /3
is

7a-/32 7 + 0-2/3'

afl-7

a+/3-27

obtained

by

the homographic transformation

axy + b(x + y) +

= Q.

Equation of Squared Differences of a Cubic. "We shall now apply the transformation explained in the preced42.

ing Article to an important problem,

viz.

the formation of the

equation whose roots are the squares of the differences of every two of the roots of a given cubic. We shall do this in the first
instance for the cubic
y?

+ qx + r
is

0,

(1)

in

which the second term

is

absent,

and

to

which the general

|3,

equation

readily reducible.

Let the

roots be a,
are]

7.

We

have to form the equation in y whose roots

(P-yY,

(7

-)', (-P)'-

here observe that the method of Art. 39 can be viz. the forapplied in general to the solution of this problem, the are mation of the equation whose roots squares of the for differences of every two of the roots of a given equation
;

We may

when
{y
is

the product

(ai

- a2 2 )

[y

( ai

03)'}

{y

(a,

- a 4) 2

.'.

.Jjf- (%-

$\

symmetric

of y will bo formed, the coefficients of the successive powers and may, therefore, be functions of a a a
l9
2,

3,

a<,

&c.,

expressed in terms of the coefficients of the given equation

Tn

82

Transformation of Equations.

the present instance, however, the method of Art. 41 leads more This equation readily to the required transformed equation. " " may be called for brevity the equation of squared differences of the proposed equation. Assuming y equal to any one of the
roots of the transformed equation,
e.

g.

(]3

- y) 2 , we have

also

a2 +

/3

+ 7 2 = -2?, =

aj3y

-n

The equation

(x,

y)

of Art. 41, becomes, therefore,

or

subtracting from this the proposed equation,

we

get

(y

+ ?)z-3r =

0,

QIX

=^
0.

Qr

hence the transformed equation in y

is

4q*

+ 27r z =

(2)

If

it

be proposed to form the equation whose roots are the

(a, )3,

squares of the differences of the roots

7) of the cubic
(3)
is

a 3 = 0,

we

first

remove the second term

the resulting equation

G
and the required equation
is
is

the same as the equation of squared

differences of this latter, since the difference of

any two

roots

unaltered

write

by down the

removing the second term.

required equation

We

can therefore

by putting

Equation of Squared Differences of a Cubic.

in the above.

83

The

result is

which has for roots

2

(/3-r)

(y-*)'.

(-/3)'.

The equation
x*

(4)

can be written in a form free from fractions

by multiplying the
+

roots

by

a
z

z
.

It becomes then
z

1SH? + SlH x + 27 (G
2 2
2

+ 4J? 3 ) =
2

0,

(5)

whose roots are

a (-/3) Q3- 7 (r-a) can write down from this an important
0o
)
,

2
.

We

function of the

roots of the cubic

rences, in

(3), viz. the

product of the squares of the diffe:

terms of the coefficients

6

0o

O - y)

(y

a) (a

]3)

= - 27 (<? +

4#

).

(6)

evident from the identity of Art. 37 that contains a<? as a factor. have in fact
It
is

+ 47p

We

The
cubic,

expression in brackets
is

is
;

called the discriminant of the

and

represented

by A
2

giving the identities

HI - a J^ A.

EXAMPLES.
JL.

Form

the equation of squared differences of the cubic

AM.

x*

-42x? + 441* -400 =

0.

L^l

Form

the equation of squared differences of

First

remove the second term.

Ana. z3_30s2 + 225s -68 = 0.
the equation of squared differences of

3.

Form

=0.

AM. x*conclusion with respect to the roots of the given cubic can be drawn from the form of the resulting equation in the last Example ?
4.

What

G2

84

Transformation of Equations.

43. Criterion of the Nature of the Roots of a Cubic. "We can from the form of the equation ^of differences obtained in Art. 42 derive criteria, in terms of the coefficients, of the na-

For ?< if the equation ture of the roots of the algehraical cuhic. has a the 42 cuhic Art. of root, negative ((3) Art. 42) must (5)
have a pair of imaginary roots, in order that the square of their and if (5) has no negative root, difference should he negative
;

the cuhic
of (3)

(3)

has

would In what

pair of imaginary roots give rise to a negative root of (5). follows it is assumed that the coefficients of the
:

all its roots real, since \ a

equation are real quantities. Four cases may he distinguished 2 + 4j5P is negative, the roots of the cubic are all (1). When G S to make this negative must he negative (and real. For,

4H

> G ) the signs of the equation (5) are [then alternately positive and negative, and, therefore (Art. 20), (5) has no negative root;
z
;

and consequently the given cuhic has

(2).

all its roots real.

When G z +

4jET

is

positive, the cubic* has two

imaginary

roots.

For the equation (5) must then have^a negative root. 3 = For 0, the cubic has two equal roots. (3). When G* + 4J7 In this case the equation (5) has then one root equal to zero. A = 0, it being assumed that # does not vanish. We may say,

therefore, that the vanishing of the discriminant (see Art. 42) expresses the condition for equal roots.
(4).

When
also

6r

0,

and

H - 0,
a

the cubic\has its three roots equal.

For the
tions

roots of (5) are then all equal* to zero.

These equa-

may

be expressed, as can be easily seen, in the form

a\

a2
#3

which relations among the

44.

coefficients'are therefore the conditions

that the cubic should be a perfect cube.

The general aid of symmetric functions, of problem the equation whose roots are the differences, or the squares of the differences, of the roots of a given equation, may be treated as
of the formation,

Equation of Differences in General.

by the

follows

Let the proposed equation be

Equation of Differences
Substituting x + a r for
/?,

in

General
1,

85
2, 3,

x,

and giving r the values

in succession,

we have the equations

- a 3 .... )

- a 3 ....

f(ar)

Also, employing the expansion of Art. 6, and observing that = 0, we find the equation

/(* + a r) -/(,) +

j/"(r)

r3 /"K)

.....

Denoting the second side of

this equation
(1),

by $ (#,
obtain
2

a r ),

and

multiplying both sides of the identities

we
2
j

(x, ai)

(a, a z )

----

OP,

an ) =

{^ -

( Ql

a 2)

[a;

(d - a 3 )*}

. .

the equation of differences, therefore, we can multiply together the n factors 9 (ar, a x ), (x, a 2 ), &c., and substitute for the symmetric functions of the roots which occur in the pro-

To form

duct their values in terms of the

coefficients.

Or we may,

as

already explained in Art. 42, form directly the product of the \n (n -I) factors on the right-hand side of the above identity,

and express the symmetric functions involved coefficients. The roots of the resulting equation

in terms of the

of the n (n

-l)

tk

degree in x are equal in pairs with opposite signs. Since the variable in this equation occurs in even powers only, we may
substitute

x for
th

\n\n

l)

and thus obtain the equation of the roots are the squared differences. whose degree
or,

third degree the formation of the shall give the equation of differences becomes laborious. result in the case of the general algebraic equation of the fourth degree in a subsequent Chapter.

For equations beyond the

We

Transformation of Equations.

EXAMPLES.
.

The

roots of the equation

are o,

ft,

form the equation whose

2

roots are
2
>

+ 72

2 7 +

Ans. y3
2.

28y

245y

650

0.

The

roots of the cubic

y?

2x*

3x

are

o,

/J,

form the equation whose

roots are

Ans. y*
3.

12y

I72y - 2072

0.

The

roots of the cubic

*3
are
a,
)8,

qx

form the equation whose

roots are
2
,

a2

aj8

2
.

4.

The

roots of the cubic

z 3 + px*
being
a, ft,

+ qx +

form the equation whose

roots are

r2
5. If a,
/3,

= 0.

7 be the

roots of the cubic

s3 - 3 (1
prove that
(

+ a + a2 x 4- 1 + 3a +
)

3a2

+ 2a 3 =
a.

- 7)

(7

o) (a

)8)

is

a rational function of

Ans.
6.

Find the

relation

between
a c z*

and
2

H of the cubic
+
Zazx
2
)
,

3ais

+ a3 =
2

when

its roots
-

are so related that (0

-7

(7

a)

(a

2
13)

are in arithmetical

pr greSSi0n
7.

.**.*
/3,

+MP-0.

f o,

7, 5

be the roots of
c 2 **

- 2<r^ 3

-f

2* 1

=
2

0,

find the value of

2

(0

- T2 2 (a* )

S2) 2

(7*

2 2
)

(/3

- S2

(cf

2 2
]8
)

(7

52 ) 2

Am.

Examples.
8.

87
+ 78 =

Prove that,

if

&y + 7 +

a/3

+ +

08

0,

= 18 { GB - 72 2 (' )

2 2
)

(7

- a)
9*
*

OS

- V)* + (a* - 0')(73 =

0,

9.

Solve the equation

x*

- **
1

-f

8* 2

15

which has one root of the form

Diminish the roots hy
o*
1
;

+ a \/~
;

substitute o
-f

v/-

for
2.

-we find that a

must

- 3a2 - 4 = 0, and a 4 other factors are (x

satisfy

6a2
1)

hence o
3), as is

Hence

the factor s2

- 2x + 5.

The

and (x1 -

evident.

10.

The

roots of the cuhic

are a,

/3,

form the equation whose

roots are

This question has been already solved in Art. 41.

tion which,

We give here another solu-

although in this particular instance it is not the simplest, will be found convenient in many examples. Let the roots of the given equation be diminished by h. The transformed equation is (Art. 35)
a y3
-I-

3^fiy
A.

+ 3A z y + A* =

0,

whose roots are a

Art. 24)

A,

/3

h,

y-

We express the
=
0.

should have two roots equal with opposite signs.

condition that this equation This condition is (see Ex. 17,

This equation

is

a cubic in h whose roots are

for the above condition is

(3

A)

(7

A)

0,

2A
where
th
/5,

7,
roots.

represent indifferently

any two of the

From
by
2.

the equation in A

required cubic can be formed by multiplying the roots 11. The roots of the biquadratic

ao*4
are
o, (3,

4fliz*

6a 2 z2

+ 4a 3 x +

a*

7, 5

form the sextic whose

roots are

7,

7+

a,

-f (3,

8,

5,

7+

5.

Employing the method of Ex. the condition of Ex. 20, Art. 24.

10, the required equation

can be obtained from

The

condition

is

in this case

AiA*Az- A^AtThis
is

aQ

AJ=

0.

a sextic in A whose roots are

(/3

7), &c.,

from which the required

equation can be obtained as in the last example.

88

Transformation of Equations.

*12. Form, for the cubic of Ex. 10, the equation whose roots are
/By
/8

-a 2
2u
h,

ya-

a)3

- y*
.

+y-

7 + a - 20*

+ ft-2y'

Diminish the roots by

hould haye
dition is
its

roots in geometric progression (see

and express the condition that the resulting cubic Ex. 18, Art. 24). The con-

This will be found to reduce to a cubic in h

written, since

whose roots are the values aboye

13.

A), or

= -^ /3 + 7

2a

Form

for the
2)37

same cubic the equation whose roots are

ajB

- ay
2a~
'

2ya -

(3y

(3a
J

2aj8

- ya - y&
+
/3

/3 -f

y+

a-2

- 27

should have

Diminish the roots by h, and express the condition that the transformed cubic its roots in harmonic progression (see Ex. 19, Art. 24). We have

a-

0-h

207 -a&-ay

/j+r-2..The equation
in A
is

0,

which
14.

will be

found to reduce to a cubic.

The

roots of the biquadratic

are o,

7, 5

find the cubic

whose

roots are

aS

70 -

- yS

Diminish the roots by

condition
is

h,

and employ the condition of Ex. 22, Art.

24.

The

in this case

=
which reduces to a cubic in h whose
15.

0,

roots are the values above written.

Find the equation whose

roots are the ratios of the roots of the cubic

x3

qx

0.

The

general problem can be solved

equation, and p

=- =

by

elimination.
;

Letf(x)

=
=

be the given
0,

the ratio of two roots

;

then since /(/3)

is

we have

f(pa)

0, also /(a)

and the required equation in p

obtained

by eliminating

Examples.
a between these two latter equations.
result is

89
in the present
0.

For the cubic

3

example the

r
16. If o,
,

(p

I)

-f

0V(p +
r

1)2

7 be the

roots of

z3

-f

pxz + qx +

0,

form the equation whose

roots are

jin*.

^ - 2 (p - 2q) xz + (p* - 4p 2 ? + 5?2 - 2^r) * Form

for the

p* q~

-2p3 r + 4pqr - 2? 3 - r2 = 0.
)

17.

same cubic the equation whose roots are

7
Ans.
r~ x*

V
x3

.7

7, +
a
(

a
'

afi
'

7
p*r

-(pqr- Sr2 ) x* +

- bpqr -f (p q*
z

Sr2

+ q*) x

2 3 -2p*r + 4pqr - 2? - r ) = 0.

18. If o,

7 be the

roots of the cubic

qx

0,

form the equation whose roots are

la

m(3y,

1(3

+ mya,

ly

ma/3.

Ans. y 3 - mqy z + (Pq + Zlmr) y +

19. If a,
(3,

Pr- Pm q* - 2/m 2 qr - m3 r* = 0.
=

y be the

roots of the cubic

aox3
find the equation

+ Zaix2 +

3a z x

0,

whose

roots are

20.

Form, for the cubic of Ex.

19, the equation

)2( 2

whose

roots are

08- 7 )2(2-0- 7 )2,

rences of the cubic

7-

0_ 7 -a)

2
,

(a-j8)~(2 7

--)8) 2

The required equation can be obtained by forming the equation

(4)

of squared diffe-

of Art. 42, since

21.

Form,

for the cubic of Ex. 16, the equation

whose
2

roots are
-

(/3-7)

2
>

0(7-)

2
,

7(-0)

Let the transformed equation be x3

+ Pz* + Qx + H =

0.

Ans.

3 P=pq-9r, Q = q - 9pqr + = - r (4? s -f 27 r 2 + 4/> 3 r -p*q* -

ISpqr).

22.

Form,

for the

same cubic, the equation whose

a2

roots are

2#y,

2
/3

+ 2 7 a,

2 7 + 2j8.

Ant.

P=-p*,

Q = q(2p z -Zq),

-& =4p*r- IBpqr

CHAPTER

V.

SOLUTION OF RECIPROCAL AND BINOMIAL EQUATIONS.

45.

Reciprocal Equations.

It has been

shown

in Art.

32

that all reciprocal equations can be reduced to a standard form, in which the degree is even, and the coefficients counting from

now the beginning and end equal with the same sign. proceed to prove that a reciprocal equation of the standard form can always be depressed to another of half the dimensions.
.

We

Consider the equation

l

+ a m xm +

+ a^x + a =

0.

m and Dividing by x uniting terms equally distant from the we have extremes,
,

aXm l+
~

Assume x + - = 2, and

let

xp +

be denoted for brevity by

Vp

We have

plainly the relation

Fjm =

Vp z - Vp

.,.

Giving^? in succession the values

1, 2, 3, &c.,

we have

F
and
so on.

- 5s 3 + 5s

Substituting these values in the above equation,

;

we

th get an equation of the m degree in z and from the values of 2 those of x can be obtained by solving a quadratic.

Examples.
EXAMPLES.
1.

91

Find the

roots of the equation

z6 +
Dividing by x

x*

+
x*

sc

+
x*

x*

+ x+l =
1

Q.

1 (see Art. 32),

we have

= 0.
=
1,

This equation

may be

depressed to the form

s2

0,

giving

whence

+ -=1,
x

+-=
x

and the

roots of these equations are

1

^/ri

-1 j-y^-3

2.

Find the

roots of the equation

Dividing by

a;

1,

which may be done

briefly as follows (see Art. 8),

1_3 5-5 3-1 1-2 3-2

-2
we have
the reciprocal equation
S

-2
- 2*2 +
1

0,

*8

2z 6

3s*

0,

(1)

Substituting for

F4

z4 z4

4z2

+
9
3,

and for
0,

F2
2
(z

s2

2,

we have
0,

the equation

6z2
s2

or
z

3)3,

whence

=
a;

and

=+v

giving

= v/3, * + - = - \/S

and the roots of these equations are

These roots are double roots of the equation

3.

(1).

Solve the equation

**
1

0.

Dividing by x

we have

from which we obtain

2

+t-

0.

92

Solution of Reciprocal
Solving this equation,

and Binomial Equations.

we have

the quadratics

=0,

from which we obtain

where
2

1.

This expression gives the four values of

4.

x.

Find the quadratic

factors of

*6
Transforming
this,

we have
2s

3*

= =

0,

whence

0,

and

\/3.

The

quadratic factors of the given equation are, therefore,

z2
\jb.

+1 = 0,

x2

<v/3 x

0.

Solve the equations

(1).

(1

+ *)* =

(1

+ * 4 ),

(2).

(1

*)

(!

/6. Reduce to an equation of the fourth degree in

(1

*)

+
,

(I-*) -

2fl

^n*.

(1

a)z*

(7

3a)

(4

a)

0.

In

General Properties. and the following Articles will be proved the leading general properties of Binomial Equations. PROP. I. // a be an imaginary root of xn - 1 = 0, then am
46.
this

Binomial Equations.

also will be a root,

being any integer.

Since a

is

a root,
1,

an =
that
is,

and therefore
am
is

n
(a
}

m =
n

1,

or (a m ) n = 1
0.

a root of x - 1 =

The same
this case

equations
unity.

0, except that in be an odd integer. 47. PROP. II. If m and n be prime to each other, the xm - 1 = 0, xn - 1 = have no common root
is

true of the equation xn

+ 1 =

m must

except

Binomial Equations.

93

To

prove this
:

we make

numbers and b can

is

If m and n

be integers

use of the following property of prime to each other, integers a

be found such that

mb -

na =
is

1.

For, in fact,

when

turned into a continued fraction,

the approximation pre-

ceding the final restoration of

Now,

if possible,

let a

be any

common
1
1

root of the given

equations; then
am =
also
1, 1,

and a n =
and a
or a
11

a a
(

mb

=
=

na

=
1,

whence
that
is,

mb ~ n ^

1,

or a

= 1

1 is the only root

common

to the given equations.

48. PROP. III.

integers

If Jibe the greatest

common

pleasure of two
1

and x

and n, the roots common to the equations xm - 1 = 0. = 1 0, are roots of the equation a?
this, let

0,

To prove
Now,
a

m = km',
since

n =

kn'.

may

prime to each other, integers 1 hence be found such that m'b - n'a =
n' are
;

m' and

and

mb - na =
If, therefore,

k.

a be a

common
=

root of xm O r a*

- 1 = 0, and^ -1 =
1

0,

(m&~na)

=I

which proves that a 49. PROP. IV.

is

a root of the equation a* - 1 = 0. When n is a prime number, and a any

in the series 0, all the roots are included

n I = imaginary root ofx

1, a,

2
,

... a-

all roots of the equaI., these quantities are two if different all are And they tion. for, ; possible, let any q = a , of them be equal, a?

For,

by Prop.

whence
but, by PROP.
sarily

(p 9)

"

>

since n is necesII., this equation is impossible,

q),

prime to (p

which

is

a number

less

than

n.

94
50.

Solution of Reciprocal

and Binomial Equations.

a composite number formed of the

PROP. Y.

Wlien n

is

p factors PJ q, r, &c., the roots of the equations x

- 1 =
0.
;

0,

xq - 1 =

0,

x r

1 =

0, &c., all satisfy the equation

xn - 1 =

p - 1 = For, consider a root a of the equation x from which we derive

(

then cf =

apyr

= i

or a 71

- 1 =

which proves the proposition. When n is a composite number formed of the 51. PROP. YI. n are 1 = prime factors p, q, r, &c., the roots of the equation x
the n terms of the product
(1

+a+

+aP-') (1

+]3+

1 .
. .

+/3*-

(1

+7+

.. r

+ 7 r -')

where a

is

We prove this
cn

a root ofxp - 1 = 0, /3 ofx q - 1 = 0, 7 of x - 1 = 0, &c. for the case of three factors^, q, r. A. similar

a 6 proof applies in general. Any term, e. g. a )3 7, of the product n an = 6n = = 1 is evidently a root of the equation x 0, since a 1, 1, j3

1,

and, therefore, (a/3 7

c
)

1.

And no two
a
b c

terms of the

product can be equal ' ' 6/ another term a a |3 y c

ber of this equation member is a root of xqr - 1 =

; for, if possible let a fl y 'c then a a a = $>-*' '. ; p = x is a root of 1 0, and

be equal to
first

y-

The

mem-

the second

equations cannot have a common root since p and qr are prime to each other hence a a j3V cannot be equal to a 0/ /3 6 V'. (Prop. II.)
0.
;

Now these two

52.

PROP.
,

YIL

The

roots of the equation

xn - 1 =

0,

where

n = p a q b rc and p,

q, r are the prime factors ofn, are the n products of the form afiy, where a is a root ofx?" = 1, /3 a root of x^ = 1,

and yofx rC = 1. This is an extension of Prop. YI. to the case where the prime factors occur more than once in n. The proof is exactly similar. n n n Any such product a]3y must be a root, since a = 1, j3 = 1, y = 1,

and a proof similar to that of n being a multiple of p a q b r Art. 51. shows that no two such products can be equal, since
,
, ;

a
,

r c are

prime to one another.

We have,

for convenience,

stated this proposition for three factors only of n. proof can be applied to the general case.

similar

From

this

and the preceding propositions we are now able

:

to derive the following general conclusion

The Special

ri

Roots of Unity.

95

The determination of the n th roots of unity is reduced to the case where n is a prime number, or a power of a prime number. 53. The Special Roots of the Equation x" - 1 = 0.

Every equation x to any equation of

similar

has certain roots which do not belong form and lower degree. Such roots
th

roots of unity. If n be a prime number, all the imaginary roots are roots of this th If n = p a where p is a prime number, any n root of a kind.
,

we

call special roots* of that equation, or special n

lower degree than n must belong to the equation xpa a~ a since every divisor of p is a divisor of p (except n
l

-1

0,
;

itself)

hence there are p a ( 1

If, again,
,

roots
j

which belong to no lower degree.

n = p a q b where p and q are prime to each other, there

are
1

p
-

a
f

1
J,

and q b f

1
j

special roots of

xP

- 1 =

0,

and

vfl

1 = 0, respectively.

roots of these equations, a/3

m = 1, not, suppose (a/3) a m a m = fi-m k u a Jg a TQQ ^ Q f xp _ I = (^ an(J fi-m [ s a roo ^ Q f xqb - 1 = 0, and these equations cannot have a common root
;
.

if a and /3 be any two special a special root of x n - 1 = for if where m is less than n we have then

Now,
is

{;

other than

1,

as their degrees are prime to each other

less

conse-

quently
x"

m
0.

cannot be

than

n,

and

a]3

is

a special root of

- 1 =

Also, as there are

PJ

37

PJ

th such products, there are the same number of special n roots. This proof may be extended without difficulty to any form of n.

All the
where a
is

all roots.

therefore a

When
*

, equal not a special n th root, since p - q is less than n. one special n th root a is given, we may obtain all the other
; ;

2 n a"' ofx -l = Q are given by the series 1, a, a 2 th that it is a For a, &c., are plain any special n root. ? = = 1 and a And no two are for, if a?

roots

a^

is

th roots of unity. special n

" here used in preference to the usual term priof numbers. in the mitive root," since the latter has a different signification theory

The term "

special root

"

is

96

Solution of Reciprocal
Since a
is
th

and Binomial Equations.

the roots
1, a,

a special root,

all

2
,

...

a"' 1 are select

different n

roots, as

we have

just proved;

and

if

we

root a? of this series, where

a",

p
.

is

prime to

n,

the roots

*,

a(n^P, a n ? (= 1)

are all different, since the exponents of a when divided by n give different remainders in every case ; that is, the series of numbers
0, 1, 2, 3, ...

n - 1 in some order

whence

this series of roots is

the same as the former, except that the terms occur in a different To each number p, prime to n and less than it (1 inorder. th mp cannot cluded), corresponds a special n root of unity for a be equal to 1 when m is less than n, for if it were we should
;

have two roots in the

all

give binomial equation of a degree inferior to n that is, ap is a special n th root of unity. What is here proved agrees with the result
:

series equal to 1, and the series could not the roots in that case ; therefore ap is not a root of any

above established, since the number of integers

less

than n and
jfl
)

prime

to it

is,

by a known property
,

of numbers,

nil

when n = p

which is also, as above proved, the number of q n x of 1 = 0. roots special

EXAMPLES.
1.

To

determine the special roots of #6

0.

Here, 6
jc

= 2x3.
3

z Consequently the roots of the equations x

1
2

0,

and
1;

are roots of x 6
a-

=0.
x

Now,

6 dividing x

3 by #

we have x3 +

and dividing *

by
a;

-, or
C
..1

1,

we have

xz

x +

0,

which determines

the special roots of

=0.

Solving this quadratic, the roots are

also since

aai
01

=
,

a6

a5

which may he easily The special roots

verified.

are, therefore,
a,

a6

or oi 5 , 01

or a, -.

Examples.
'1.

97

To

discuss the special roots of

a;

12

0.

Since 2 and 3 are the prime factors of 12, and

1

=
2

6,

=
3

4,

the roots of

= 0,
4

and x* -

and x 6
of x l -

1,

+1= - I =
+
1

1 by z* - 1, and equating the quotients to zero, we have the two equations 6 0, ami .- + 1=0, both of which must be satisfied by the special roots

0, are roots of

a;

12

now, dividing x l ~

therefore, taking the greatest

it

common measure

of

& + x* +

1,

and

z6
x*

+
-

1,

and equating

to zero,

the special roots are the roots of the equation

12 1 by the dividing a: the reciprocal equation

3?

0.

The same result would plainly have been arrived at by common multiple of z4 - 1 and x* - 1 Now, solving
.

x*

- x" + 1 =

0,

we have x + - = + v^3

whence,

if

a and

ai

be two special

roots,

re the four special roots of

a;

12

0.

"We proceed now

Since

to express the four special roots in terms of

any one of them

a.

4-

ai

+
ai

0,

or (a

ai)

0,

'\

aai)
assigned to a and ai)
5
;

we

take aai

1 (as

consistent with the values

we have
1,

and
roots

a and ai are roots of 3?

0,

=-

and a =

-- =
a
,

<n.

The

a, ai,

ai

mav therefore be
5

5 expressed by the series a, a , 7 11

a",

a 11 since a 12 =

1.

Further, replacing a by a , a , a , we have, including the series just determined, the four following series, by omitting multiples of 12 in the exponents of a
:

a,
a5,
a",

a5
a,

a7

a 11
,

a 11
,

a7

a 11
,

a,

a,
,

11

a7

a5

a;

their order only where the same being changed. "We have therefore proved that this property is not peculiar to any one root of the four special roots ; and it will be noticed, in accordance with what roots are reproduced in every
is

row and column,

and

above proved in general, that 1, 5, 7, and 11 are We may obtain all the roots of x l ~ less than it.
a,

all
1
:

the numbers prime to 12, = by the powers of any

one of the four special roots

a 5 a7 a 11 as follows
,
, ,

98
3.

Solution 'of Reciprocal

Prove that the special roots of # 15 x8

and Binomial Equations.

1

are roots of the equation

x~

x5

x*

xz - x

+
1

0.

4.

Show

that the eight roots of the equation in the preceding example

may

he

obtained by multiplying the two roots of #2

x*
-i
a;

+x+
1

by the four
0.

roots of

x*

+x+

54. Solution of Binomial Equations by Circular Functions. We take the most general binomial equation

xn = a + b v/-

1,

where a and

b are constants.

Let
then

R cos a, b = R sin a xn = R (cos a + v/- 1 sin a)

a =
;

now,

if

r (cos 9 +

^/-

1 sin 9)

be a root of this equation, we have, by

r (cos n9 +
n

De

Moivre's Theorem,
sin a)
;

*/-

1 sin n&)

R (cos a +.\/- 1

and, therefore,

rn cos n9 =
r
n

R cos a, sin n9 = R sin a.

and adding,
n = giving r

Squaring these two

equalities,

r zn =

R\

where we take

and

r both positive, since in expressions of the

kind here considered the factor containing the angle be taken to involve the sign.

may

always

We have then
cos n9 = cos a,
sin

n9 =

sin a

and, consequently,

n9 = a +
k being any integer
general type
n/r> R V
cos
r ;

2&7T,
is

whence the assumed n th root

of the

+ v/- 1

/i

em

Giving to k in

this expression

any n consecutive values in the

Solution ty Circular Functions.

series of

numbers between - oo and + GO we get and no more than n, since the n values recur in
,

all

the n th roots

periods.

"We
form

may

write the expression for the n th root under the

a / cos- + v/-

sm -

n\

cos

-n

7r

+ </ - 1 sin si
1

If

we now suppose JR =
n

1,

and a =
;

0,

the equation

of

=a +b

becomes x =
root of 1 +

+
1,

*/-\

the general type, therefore, of an

is

v/-

or unity,
cos

n
\

2A-7T

-V/

/-

2A-7T

1 sin

If

we

give k any definite value, for instance zero,

E cos - + V/TT/ n
is

A/-

sm

b </- 1. The preceding formula shows, therefore, that all the n th roofs of any imaginary quantity may be obtained by multiplying any one

one n th root of a +

th of them by the n roots of unit//. Taking in conjunction the binomial equations

xn = a +

v/ -

1,

and xn = a -

*/-

1,

we

see that the factors of the trinomial

x*
n

are

R V/T>

cos

+ 2 ^ 7r

2R

cos a

xn +

v/ 3 ...

sm

where k has the values

0, 1, 2,

1.

100

Solution of Reciprocal

and Binomial Equations.

EXAMPLES.
1. Solve the equation x 1, this Dividing by x
1

1
is

=0.

reduced to the standard form of reciprocal equation.

Assuming

x+

-,

we

obtain the cubic

from whose solution that of the required equation 2. Resolve (x + 1)7 - x 1 - 1 into factors.

is

obtained.

Ans. 7z(z
3.

Find the quintic on whose solution that of the binomial equation x 11

Ans.
z5

depends.

+ z*- 4z 3 -

3z 2

+ 3z + 1 = 0.

4.

When

a binomial equation
1,

is

equation (by division by x has all its roots imaginary.

5.
3

1,

reduced to the standard form of reciprocal or #2 1), show that the reduced equation
15, 16, p. 33.)

(Cf.

Examples

When
;

this reduced reciprocal equation is transformed

by the

substitution

=z+-

show

that the equation in z has all

its

roots real,

and situated between

2 and 2.

For the
Art. 54)
;

roots of the equation in

is

z are of the form cos a +

a,

\/-

sin a (see

hence z + z
2 and 2.

of the form 2 cos

and the value of

this is real

and be-

tween
6.

t
it
:

Show

that the following equation is reciprocal, and solve

(z*

-z+

I)

27x*(z

I)

0.
2, 2,

Ans. Roots
7.

,.,-

1,

1.

Exhibit

all

the roots of the equation z

0.

The

solution of this is reduced to the solution of the three cubics

#3 where w,
o>

0,

#3

oa

0,

z3

&)

are the imaginary cube roots of unity.

:

The nine
u*.

roots

may

be repre-

sented as follows

1, 2

o>3,

o>5,

(a,

w'i,

coi!,

o>

a)?;,

the other Excluding 1, o>, o> and are the roots of the sextic
;

six roots are special roots of the given equation

.r

6.

8.

th Reducing the equation of the 8 degree in Ex.

3,

Art. 53,

by the

substitution

= x +

_
}

We

obtain

Examples.
prove that the roots of this equation are
2ir
4ir

101

Sir

14ir

cos-,

cos-,

cos-,

2 cos

9.

Reduce the equation

4*4

- 85^ + 357s2 - 340* +

it.

64

to a reciprocal equation,

and solve

Assume
10. Solve the equation

x 2 = -+ -

Ans. Roots:

-},

1,

4, 16.

x*

+ mpx3 + nPqx * +

nflpx

-f

w4 =

0.

Dividing the roots by m, this reduces to a reciprocal equation. 11. If o be an imaginary root of the equation x n - 1 = 0, where

is

a prime

number

prove the relation

(1

o) (1

o2 )

(1

o3 ) ...

(1

o'- 1 )

M.

12.

Show
Show

that a cubic equation can be reduced immediately to the reciprocal

18, Art. 24, exists

form when the relation of Ex.

13.

amongst

its coefficients.

that a biquadratic can be reduced immediately to the reciprocal form

its coefficients.

when

the relation of Ex. 22, Art. 24, exists amongst

14.

Form

the cubic whose roots are

fi

o3
1

+ =

o4
0.

o2

o5

where o

is

an imaginary root

of x 1

Ans.
15.

Form

the cubic whose roots are

o + o8

+ o 12 + o 5

a2

+ o 3 + a 11 + o10
1

o4

-f

a6

a9

+ a7

where a

is

an imaginary root of #

13

= 0.
Ans. x3

+ x"- -

0.

16. If 01, o 2 , 03

... on be the

roots of the equation

show how

to

form the equation whose roots are

1 1
,

1
,

01

+
01

02

+
02

+
a/ t

"We have here the identity

x" + p\ x*- 1

+p z x-2 +.. +p n .i x +p n = (x.

01) (x

- o2

(x

- on)

and changing x into -

(see Art. 32),

102

Solution of Reciprocal

and Binomial Equations.

,

M Multiplying together these identities, and dividing hy #

the factors on the

and assuming x + - = z, the leftX X \ O/ hand side can he expressed as a polynomial of the n th degree in z by means of the relations of Art 45.
right-hand side take the form x
4I

--

+-

) ;

17.

Find the value of the symmetric function 2a2 j8 2 (7

8)

of the roots of the

equation

This can he derived from the result of Ex.

19, p. 52,

hy changing the

roots into

--o (1
2 by o

1\
)

&/
o

of the transformed equation, and multiplying

52

which

is

equal to ^T.

Ans.

2o2 )8 2 (y -

5)

= 48

2
(

*)

From the values of the symmetric functions given in Chapter III. several others can be obtained by the process here indicated. 2 2 2 18. Find the value of the symmetric function 2(oi a n z of the a2) aa a4
.
. .

roots of the equation

a xn

-^+ naix- + n(w-l)

1

a%xn

-z

+ nan -\x + a n =

0.

We

2 easily obtain #o 2(ai

o2 ) 2

= w2

1) (ai

2)

and changing the roots

into their reciprocals

flo

we have
o2 ) 2 as 2 a4 2
. .

2 (ai -

2
tt

(n

1) (a n .r

- fl-8

n) .

19.

Show

that the five roots of the equation

x5

bpx*

5p*x

are

%/a+%/b,

ef/

where ^/ ab

p, a

q,

and

Q is

an imaginary

fifth root of unity.

N.B.

form can consequently be immediately solved. 20. Form the biquadratic equation whose roots are
quintic reducible to this

+ 2o4

o2
5

2a 3
1

o3
0.

2a2 ,

o4

2a,

where a

is

an imaginary root of x

Ans. * 4

3* 3

x"-

3*

11

0.

CHAPTER

VI.

ALGEBRAIC SOLUTION OF THE CUBIC AND BIQUADRATIC.

the Algebraic Solution of Equations. Before proceeding with the solution of cubic and biquadratic equations
55.

On

introductory remarks, with a view of putting the student the general principles on which the before clearly of these equations depends. With this object solution algebraic we give in the present Article three methods of solution of the
quadratic, and state as we proceed how these methods may be extended to cubic and biquadratic equations, leaving to subse-

we make some

quent Articles the

involved.
(1). First

complete

development of the principles

:

its

by resolving into factors. Let it be required to resolve the quadratic xz + Px + Q into For this purpose we put it under the form simple factors.
x~

method of solution

Px + Q + Px + Q

9 -

0,

and determine

so that

x2 +

may

be a perfect square,

i. e.

we make
p2 _

p2

4Q

whence, putting for 9

tf

its

value,

we have

PX + Q

v/P^
to the

its

Thus we have reduced the quadratic - v. simple factors are u + v, and u

form

u"

- v2

nd

104

Algebraic Solution of the Cubic and Biquadratic.

Subsequently we shall reduce the cubic to the form or u z - v z (ix + m)* (I'x + m)*,
,

and obtain

its

solution from the simple equations

u-v=
It will be

0,

u-

CDV

0,

u - w zv =

0.

shown

also that the biquadratic

may be

reduced to

either of the forms

(W + mx +
z

n)

(l'x

mx + n'}\
q'),

(x

+ px +

q} (x
;

+ p'x +

by solving a cubic equation and, consequently, the solution of the biquadratic completed by solving two quadratics, viz., in the f 2 first case, Ix* + mx + n = and in the second case, (I'x + m'x + n )
;

x*

+px + q =
(2).

0,

and xz +p'x + q' =

radicals.

0.
:

Second method of solution

by assuming for a root a general

form invoking

Since the expression p + ^/q has two, and only two, values when the square root involved is taken with the double sign, this
is

a natural form to take for the root of a quadratic.

Assuming,

therefore, x

= p + */q, and rationalizing, we have z x* - 2px + p - q = 0.

2 equation be identical with x +

Now,

if this

Px + Q =

0,

we have

2p=-P, p*-q=Q,
giving

which

is

In the case

the solution of the quadratic equation. of the cubic equation we shall find that

Vp +

Vp

rr=,

and %/p

q (v/p +

\/q)

are both proper forms to represent a root ; these formulas having each three, and only three, values when the cube roots involved

are taken in

In the

all generality. case of the biquadratic equation

we

shall find that

Algebraic Solution of Equations.

105

are forms which represent a root ; these formulas each giving four, and only four, values of x when the square roots receive
their double signs.
(3). roots.

Third method of solution

by symmetric functions of the

Consider the quadratic equation xz + have the relations roots are a, /3.

Px + Q =

0, of

which the

We

a +

j3

= - P,
Q.

a/3=
If
fall

we attempt to determine a and j3 by these equations, we back on the original equation (see Art. 24) but if we
;

could obtain a second equation between the roots and coefficients, of the form la + mfi =/(P, Q), we could easily find a and /3 by

means

of this equation

and the equation a +

]3

= - P.

in the case of the quadratic there is no difficulty in the required equation; for, obviously, finding
(a

Now

j8)

- 4Q

and, therefore, a
+

13

=
2

v/P^4Q.

In the case

of the cubic equation #3

Px + Qx +
jR),

R = 0,

we

require two simple equations of the form

la

mp + ny=f(P,

Q,

in addition to the equation a + j3 It will subsequently be proved that the functions roots a, /3, y.
(a

+ y = - P, to determine the + w 2/3 + wy) 3

+ u)3 +

a>

y)

(a

may
ratic

be expressed in terms of the coefficients by solving a quad;

equation and when their values are the cubic may be easily found.

known

the roots of

In the

case of the biquadratic equation

0*

+ P0 + gr' + Jto +

-0

we

require three simple equations of the form

la
-f

w/3 + ny + rS =/(P, Q, E, 8),

in addition to the equation

106

Algebraic Solution of the Cubic and Biquadratic.

j3,

to determine the roots a,

y, S.

It will be proved in Art. 66,

that the three functions

may

be expressed in terms of the

;

coefficients

by solving a

cubic

equation

and when

their values are

known

the roots of the

biquadratic equation may be immediately obtained. In applying the principles here explained to the solution of the cubic and biquadratic the order of the present Article is

not followed.

The student

will

have no

difficulty in perceiving

under which of the methods here described any such solution should be included.
56.

Tbe Algebraic Solution of the Cubic Equation.

Let the general cubic equation

be put under the form

s
3

+ 3 #2 +

G = 0,
(Art 36).

where

= ax +

b,

H^ac-b\ G^cfd-3abc + 2b*

To

solve this equation, assume*

hence, cubing,
s
3

=p + q
z*

therefore

Now, comparing

coefficients,

we have

Vp \/q = -H,
from which equations we obtain

p+q=

=i (-

G
is

* This solution

usually called Cardan's solution of the cubic.

See Note

at

the end of the volume.

Algebraic Solution of the

CuJ>!<'.

107

and, substituting for t/q

its

value

we have

-H

as the algehraic solution of the equation

It should he noticed that if

3 is

he replaced

"by q this

value of
;

unchanged,

as the terms are then simply interchanged

z

also,

since

\/p has the three values \/p, w\/p, w \/p, obtained by multiplying any one of its values by the three cube roots of unity, we obtain three, and only three, values for s, namely,
/~ VP+
a

~~

T7^,.

vp

"Vp + u?

:\

-"
-_,

VP

/ o>V> + w
3

H
:;

vp

the order of these values only changing according to the cube

root of

selected.
if z

Now,

be replaced by

its

value ax + b

we

have, finally,

Vp
(where

has the value previously determined in terms of the

cubic equation coefficients) as the complete algebraic solution of the

ax z
the square root and cube root involved being taken in their entire
generality.
57.

Application to Numerical Equations.

The

solu-

tion of the cubic which has been obtained, unlike the solution of the quadratic, is of little practical value when the coefficients of

the equation are given numbers

tion
it is

although as an algebraic soluall real,

;

For,

when

complete. the roots of the cubic are

CP + 4ZP = - K\
and, substituting

an essentially negative number for p and q their values

(see Art. 43)

108

Algebraic Solution of the Cubic and Biquadratic.

in the formula \/p

+ \/#, we have the following expression

:

for

a root of the cubic

Now there is no general arithmetical process for extracting the cube root of such complex numbers, and consequently this formula is useless for purposes of arithmetical calculation.
But when the
proximate numerical value
cubic has a pair of imaginary roots, an apmay be obtained from the formula

since

+ 4ZP is positive in this case. As a practical method, however, of obtaining the real root of a numerical cubic, this
process
is

of little value.
first

In the

case; namely,

where the roots are

all real,

we

can make use of Trigonometry to obtain the numerical values of

the roots in the following

manner

Assuming

2R

cos

= - G, and 2 R sin

we have
also

p=
tanf =

-E
w=

and
TT

R=
-

2 (G +

)*

(-

and

ion.

finally, since

cos

v/- 1 sin

?r

ir

-/n

the three roots of the cubic equation

s
3

3Hz

0,

viz.

vp + ^q,

2 w\/p + w \/q,

become

from which formulas we obtain the numerical values

of the roots

Expression of the Cubic as Difference of

of the cubic

tioo

Cubes.

109

of a table of sines and cosines. This process in is not convenient practice ; and in general, for purposes of arithmetical calculation of real roots, the methods of solution of

by aid

numerical equations to be hereafter explained (Chap. X.) should be employed.

58.

Expression of the Cubic as the Difference of two

Let the given cubic

Cubes.

be put under the form

where

= ax

-f

b.

Now

assume
3

v)

-v(s + M )
:

J,

(1)

where p and v are quantities to be determined

of this identity becomes,
z
3

the second side

when

reduced,
(fi

- 3 pv z - fiv

v).

Comparing
therefore

coefficients,
JULV

= - H,

fJLV (fJL

v)

=-

G
where
also
2

= GP +

4H\
/

as in

Art 42
=
z
2

ft
(s

+ u)(s +

v)

+ -^z-J2
JLL

r
.

(2)

Whence, putting

for z its value, ax

b,

we have from

(1)

which

is

the required expression for 0(#) as the difference of

two cubes.

By

the aid of the identity just proved the cubic can be re-

110

Algebraic Solution of the Cubic and Biquadratic.

its

solved into

completed. the equation $(x) = in terms of mial cubic the equation

(fj.

We

simple factors, and the solution of the equation proceed to obtain expressions for the roots of
/u

and

v.

Solving as a bino-

z -v) a

(f>

(x)

ju

(z

v)

- v

(3

fj)

=
:

0,

we

find the three following values for z = ax

3/~ (3/ v \V V/- V

3
fj.

fjL

+v
IUL

/~\
y;>
2

V/u %/v
'

(o>V
2

o>

v)>

jUVV

(<t

/X

<i>

vJ'

replaced by any pair of cube roots selected one from each of the two series
If
3

now A//A and \/v be

/~

~
2

/~

V/tt,

tt>v/Lt,

ftrv/ttj
2

a/ v v>
it

s/ ov v,

'>

V/~ v,
3

will be easily seen that we shall get the same three values of z, the order only of these values changing according to the cube roots selected. It follows that the expression
3

V/XVy
taken in

3/~ /3/~ 3/~N \V fJL+V V)

has three, and only three, values when the cube roots therein are This form therefore is, in addition to all generality.

that obtained in the last Article, a form proper to represent a root of a cubic equation (see (2), Art. 55).

The

function

(2)

becomes, as

may

given above be easily seen,

when transformed and reduced

(ac

b*) x?

+ (ad -

be)

x + (bd -

2 )
.

This quadratic, therefore, contains as factors the two binomials ax + b + fa ax + b + Vy which occur in the above expression of ^ (x)
as the differences of

two cubes.

Solution of the Cubic by Symmetric Functions.

59.

1 1 1

Solution of the Cubic by Symmetric Functions

Since the three values of the expression
(a

of the Roots.
|
{a

+ j3 + 7 +

+ w)3 + ary) +
to,

(a+

a>

/3

(07)},

when

6 takes the values

1,

or, are a, /3, 7, it is plain that if

the functions
(a

+ w)3 + w 2 7 ),

(a

<o

/3

017)

coefficients of the cubic, we could, values in the formula given above, arrive their substituting by at an algebraical solution of the cubic equation. Now this cannot

were expressed in terms of the

be done directly by solving a quadratic equation the product of the two functions above written
;

for,
is

although a rational

symmetric function of
in question

a, /3, 7, their

sum

is

not

so.

It will be

found, however, that the

sum

of the cubes of the

is a' symmetric function of the roots, be fore, expressed by the coefficients, as we proceed to show. For convenience we adopt the notation

two functions and can, there-

L = a + o>|3 +

2
<i>

7,

M=a+w
,

/3

0)7.

We have then = (OL\* A + Bu+

where

<7o>

(0l3f)

^=a

3 3 +/3 + 7 +6a/3 y ,

=3(a

/3

+
j

from which we obtain

- 32a 2 ]3 + 12aj3 7 = - 27
(Cf.

Ex.

5, p.

44; Ex. 15,

p. 50.)

Again,

whence

(a

+ w/3 +

a>

7)

(a

a>

)3

0)7)

are the roots of the quadratic equation

z

G
a3

e^A
a6

Denoting the roots of

this equation, viz.

112
by
tl

Algebraic Solution of the Cubic and Biquadratic.

and

4, the original

formula expressed in terms of the coefthe three roots the expressions

ficients of the cubic gives for

--+a

=
a

+ 3
j3,

It will be seen that the values of a,

of the

7 here arrived

at are

same form

as those already obtained in Art. 56.

It is important to observe that the functions

(a

wj3

+w 2 7)

3
,

(a

a>

/3

+ wy) 3

are remarkable as being the simplest functions of three variables

which have but two values when the variables are interchanged
It is owing to this property that the solution in every way. of a cubic equation can be reduced to that of a quadratic equaSeveral functions of a, |3, 7 of this nature exist, and it tion.
will be

tions are connected

coefficients.

proved in a subsequent Chapter that any two such funcby a rational linear relation in terms of the
discussion of the different

Having now completed the

modes

of algebraical solution of the cubic, we give some examples involving the principles contained in the preceding Articles.

EXAMPLES.
^
1.

Resolve into simple factors the expression

(j8

-7

2
)

(x

- a) 2 +

(7

- a) 2

(x

+ (a - 0) 2

(x

-7

2
)
.

Ans.
2.

Prove that the several equations of the system

3
()8

7)

(x

a)

= (7 - a) 3 (x

3
ft}

(a

$) (x

>

have two factors common.

Examples.
Making use
whence
of the notation in the last Example,

113
we have

U* = F 3 = 7F3
)

ur+ F2 =i (usince

Z7+ F-f 7F=

therefore
is

-7 (0

2
)

(*

a)

(7

a)

(z

2
)8)

(a

2
/9)

(*

7)-

the
3.

common

quadratic factor required.

Resolve into factors the expressions

(1). (2).
(3).

OB OB (^

-7 -7
-7

3
)

(x
5

3
)

(7

a)
5
)

(* (x

0)

+
+

(a
(a

jB)>

(* (*

(*(*

5
)

+ +

(7
(7

#+
7 /8)

/B)

- 7) 3 -75
)

7
)

7
)

7
)

(*

(a

(x

- y)\
Ex. 40,

These factors can be written down

p. 59.
to,

at once

from the

results established in

Using the notation of Ex.

Z7,

1,

and replacing
:

ai, /8i, 71, in

the example referred

by
4.

F, 7F,
(1)

we

obtain the following

(2)

Ans.

3J7FJF;

|(Z7

+ F-+

W*)WW\
)

(3)

Express

(-)(jp-/i)(- 7
as the difference of

two cubes.

Assume

(*-a)(*-/B)(z- 7 )=
whence
JJi

ZTi

Fi

= *(*;-

a),

Adding,

we have
A + ^

0,

Aa + n&

vy

and, therefore,

\ = p(jB-7),

M=f>(7-a),

v=p(a

but

A/tti

whence

Substituting these values of X,

/*,

and using the notation of Ex.

2

1,

ri-Fi =
whence

pZ7;

wZ7i-

Fi

= pF,
2

CTi

-Fi = pJF;

3^i = p(J7+

F+w F+

7F),

and

Z7i
').

and Fi are completely determined.

Prove that

and Jf are functions of the differences of the

a

roots.

We have
cession,

i=

w/8

+ w2 7 =
+
2

a
;

-A+

(j3

A)

(7

A)

for all values of A, since 1

-f o>

we

obtain three forms for

= and giving to A the values a, j8, 7, in sucL in terms of the differences - 7, 7-0, a - 0.
I

Similarly for Jf.

114
6.

Algebraic Solution of the Cubic and Biquadratic.

To
express the product of the squares of the differences of the roots in terms

of the coefficients.

We have
Z + M=2a-fi-y, Z + w 2 Jf=(2j8-7-a)o>, Z +
and, again,
a>

M = (2y - a -

/3) a,

2
;

Z-Jf=08-7)(-

3
),

o>

Z- uiM= (y-o)(w3

<o

2
),

wZ - o> 2 Jf = (o -)(- w 2 ),
;

from which we obtain, as in Art. 26,

Z 3 + M* = Z3 and since
(Z
3

(2

jB

7) (2)3

-7-

)(2 7

0),

Jf 3

=-

v3
3 2
)

(0
3

-f

7) (7

(a

- I/

(Z
3

Jf
3

3 2
)

- 4Z 3 J/

3
,

we

have, substituting the values of

* e (j8

Z +
2

and ZJf obtained in Art. 59,

7)

(7

a)

(a

2
)3)

=-

27 (

4ZT 3 ).

(Of. Art. 42.)

7.

Prove the following identities

Z,
3

+ Jf = |{(2a3

jS

- 7 3+
)

(23
(7

- 7-

3
)

+ -

(27
3 /3)

}.

3) },

Z3 -

J/ 3

^^3

{(j8

-7

3
)

a)

-f

(a

These are easily obtained by cubing and adding the values of

Z + M,
in the preceding example.
,

&c.

Z - M,

&c.,

2 M"2 &c., in terms of a, ft, y. 8. To obtain expressions for Z 2 are obtained by subtracting The following forms for Z 2 and
,

M
2

(a

2
)8

+ 72

(1

+ -

&)

+ w2 =
)

from

(a

+
8

+ w 2 7) 2 and
,

(a

+ or +
,

o>7)

Z = - Jf 2 =
2

(0
(3

- 7) + -7 2+w
)

(7

s
/8) 2

(7

(a

j8)

In a similar manner, we find from these expressions

Also, without difficulty,

we have

the following forms for ZJf, and

L 2 M~:

9.

There are

six functions of the type of

Z
+

or

M,

viz.,
to
2

-I-

wj8
-

+
+

7,
017,

wo

2
-I-

7,

+ 3

-f

uy,
7,

a
to

wa + 3 +

'

<a'-y,

<a

foim the equation whose roots are these

six quantities.

Examples.
These functions may be expressed as
Z,
follow.:

\ \ :>

o*Z,
<oAT,

orZ,

M,

o-M-

hence they are the roots of the equation

(<(>

- Z)

(<J>

u,Z)

(<}>

orZ) ($ -M)(<t>3

uM

(<(>

- tfM] =

0,

or

06
Substituting for

(Z

+ jp)

4,3

L and

.If

from the equations

we have

this equation expressed in terms of the coefficients as follows

10.

To form,

in terms of

and

Jf, the equation

whose

roots are the squares of

the differences of the roots of the general cubic equation.

Let
4>

(-0) 2
u)L

hence, by former results,

V/^30 =
Rationalizing
this,

we

obtain

which

is

the required equation.

results of

In a similar manner, by the aid of the

Ex.

8,

the equation of

-quared differences of this equation, or the equation whose roots are

03-7)
i>

(2-0-7) 2
substituting

72

)-

(2)8-

7 -a)

2
,

-Wpy-a-0\
respectively, in the last
all
l.y

obtained

by

- Z and - M~

for

M and Z,

Finally, equation ; and this process may be repeated any number of times. these equations may be easily expressed in terms of the coefficients of the cubic

means of the

relations

LM=-9~, aFor instance, the

first

and

I? + J/ '=
:

27 ^. tt

equation

is

(++')*+
<Cf. Art. 42.)

i2

116

Algebraic Solution of the Cubic ar,d Biquadratic.

y and
a', #',

11. If a, 0,

7'

be the roots of the cubic equations

ax3

3bxz + 3cx + d
Sb'x*

=0,
;

ax3 +
to

Zc'x

+ d' =

form the equation which has for roots the six values of the function
<

'

oa'-f

77'.

The

easiest

mode

of procedure is first to form the corresponding equation for the

cubics deprived of their second terms, viz.,

and thence deduce the equation in the general case transformed the corresponding function
4>o

for in the case of the cubics so

(aa

b) (of a'

+ V)

-f

(j8

b)

(dff

V)

(ay

b) (a'y

Substituting for the roots of the transformed equations their values expressed by
radicals,

we have
a

v~*)

v? +

*>*

j/q
which reduces to

Cubing

this,

we

find

Now,

substituting for p and q, p' and

j',

their values given

by the equations

zz

+ Gx- H* =
^>o

0,

x~

+ G'x - H'* =

0,

we have

the six values of

given by the two cubic equations

0o

27 1IH'

c/>o

(GG'

aa

where
2

A=

G- + 4J/,

and
<w'<

'2

A' =

Finally, sulstituting for

together,
1

its

value
.

3ii',

e xz

we have the required equation It may be - 1 = 0, 0> = a + a?j8 -f w 2 7, &c., which case

noticed that

and multiplying these cubics if one of the cubio

has been already considered in

Ex.

9.

Mr. M. Roberts, Dublin Exam. Papers, 1855.

Examples.
12.

117
p,

Form

the equation whose roots are the several values of

p

where

=
p)

o -ft
.

*-*
+ py = ;

Since

a
substituting for a,
ft,

(1

y, their values in terms of p, q,

(1

and putting

\= 1 wo have

+ p)-f

par,

n=

(1

+ p)w 2 +

pa>,

Cubing, and substituting for p, q their values,

Squaring,

and by previous

results

substituting these values,

2

we have
p

the required equation

p
2 3
)

A (1 +

27J5P(p

2
)

0.

13.

Find the

relation

between the
ax*

coefficients of the cubics

3bx-

+3cx + d =

0,

when

the roots are connected by the equation

a(fi'-y')

-f

(/iJf =
'

')

y(a'

')

0.

Multiplying by

a>

a>

2
,

this equation

becomes
L' M.

Cubing, and introducing the

coefficients,

we

find

the required relation. 14. Determine the condition in terms of the roots and coefficients that
cubics of Ex. 13 should become identical
x'

the

by the
q.

linear transformation

= px +
+

In

this case

a?=pa +
Eliminating p and
q,

q,

& = p&
+ ya -

q,

y'

= py +

<!

we have
'7

7'

y'a

-f aft'

a' ft

0,

which

is

the function of the roots considered in the last example.

is

This n-lation,

moreover,

unchanged

if for o,

ft,

a,

ft',

y',

we

substitute

la

+m,

Ift

m,

ly

118 Algebraic Solution of the Cubic and Biquadratic.

whence we may consider the cubics in the
2
3

last
3

example under the simple forms

G'

+ ZHz + O =

0,

z'

+ 3.HV +

0,

obtained by the linear transformations z = ax + #, z'= a'x' + b' ; for if the condition holds for the roots of the former equations, it must hold for the roots of the latter.

Now putting tf =

kz, these

equations become identical if

'

'=

whence, eliminating

&,

is

the required condition, the same as that obtained in Ex. 13. It may be observed that the reducing quadratics of the cubics necessarily become identical by the same
transformation, viz.,

60.

Romographic Relation between two Roots of a

Before proceeding to the discussion of the biquadratic
the following important proposition relative to the

Cubic.

we prove
cubic
:

The

roots of the cubic are connected in pairs by a homographic

relation in terms of the coefficients.

Eeferring to Exs. 13, 14, Art. 27,

we have

the relations

{a ()3
f o 2

y)

j3
2

(7 (y

a)
2

+ 7

(a

|3)
2

=
}

9(a

{a

(/3

2 7) +

j3

a)

+ 7 2 (a -

/3)

= 18(^ 8

Using the notation

ffos

a\

= H,

doffs

a\ <h

2-ETi,

^2

H*

multiplying the above equations by

a)3,

(a +

)3) 5 1,

respectively,

and adding
o
3

since
/3)

-a(a +
7) (7
2

+ a|3-0,

)3

& (a

(5)

a]3

0,

we have
2 o

(j3

a) (a

]3)

18{ZTaj3 + -Hi (a +

j3)

JSij
2

but
4 o

(13

7 ) (7
;

- a) 2 (a - /3) 8 = - 27A - 108 (fiSi -

.BT, )

(see Art. 42)

whence
-

First Solution by Radicals of the Biquadratic.

and, therefore,

119

which

It is to be obthe required homographic relation. served that the coefficients in this equation involve one irrational
is

quantity, the second sign of which will give the relation between a different pair of the roots.

,61. First Solution by Radicals of the Biquadratic. Euler's Assumption. Let the biquadratic equation
ax*

+ 4fo 3 + 6cz* + 4dx +

be put under the form (Art. 37)

z*

6Hz + 4Gz +
z

a*I -

3#
G

0,

where

&

= ax +

b,

H=ac-b\ I=ae-4bd+
To

3c

2
,

= a*d - 3abc +

W.

solve this equation (a biquadratic wanting the second term) Euler assumes as the general expression for a root
z

Squaring,

Squaring again, and reducing, we obtain the equation

Comparing

this equation

with the former, we have

*/p ^/q

^r = - -

and consequently

>,

^,

r are the roots of the equation

V
or, since

~t 4/4
a*J,
etf

(1)

where

45' - (THI +
2bccl

(Art. 37),

J^ace +

- ad' -

- c\

120

Algebraic Solution of the Cubic and Biquadratic.

this equation

may
4
(t

be written in the form

HY - a*I(t + H)+a*J=Q',
t

and

finally,

putting

H= a

O,

we

obtain the equation

(2)
;

4a z O*-IaO +
This
is

J=0.
by
that name.

called the reducing cubic of the biquadratic equation

and
it is (2),

will in

what follows be referred

to

When
(1)

necessary to

make a
t

distinction

between the cubics

and

we

shall refer to the former as Euler's cubic.

Also, since

= b z - ac + a z 9

if 0i,

2,

be the roots of the

reducing cubic,

we have
q =
b*

b*

- ac + az 0i,

- ac + a z

2,

z f = b - ac

+ a2

and, therefore,

-ac + a 2 Oi + /b z -

ac

a*

62 +

\6

- ac +

a?

3.

If this formula be taken to represent a root of the biquadratic in z, it

must be observed that the

;

radicals involved

have not

complete generality would be given by the formula. is imposed by the relation

of four

for if they had, eight values of z in place

The proper

limitation

which

squaring to obtain the value of pqr) such to be attached to each of the quantities requires signs
(lost sight of in

A/r that their product may maintain the sign mined by the above equation thus,
^/P>
> ;

V^

deter-

are all the possible combinations of </p, */q,

this condition, provided that v/p, </q,

^r

retain the

*/r fulfilling same signs

throughout, whatever those signs

remove

all

be. We may, however, ambiguity as regards sign, and express in a single

may

First Solution by Radicals of the Biquadratic.

algebraic formula the four values of
s,

121
of the

by eliminating one

quantities y/p, v/q, */r from the assumed value of z by means of the relation given above, and leaving the other two quantities unrestricted in sign.

The

expression for z becomes therefore

G
a formula free from
four, values of z
all

ambiguity, since

it

gives four,

and only
:

receive their double signs the sign given to each of these in the two first terms determining that which must be attached to it in the denominator of

when */p and ^/q

the third term.

And finally, restoring

2

to p,

<?,

and

z their

values

given before,
ax + b =

we have

y/ 6

- ac + a

Oi

+ <\b~ -ac + a 2 O z

G
or 0,

as the complete algebraic

0!

solution

of the biquadratic equation;

and

being roots of the equation

4a*8*-IaO + <7=0.

To assist the student in justifying Euler's apparently arbitrary assumption as to the form of solution of the biquadratic,
we remark
absent, the
that, the second

sum

of the four roots

term of the equation in z being is zero, or Si + s 2 + z 3 + s 4 =

;

and consequently the functions

which there are in general six (the combinations of four quantities two and two), are in this case reduced to three so that we may assume
(si

s2)

&c., of

(s 2 (z 3
(*i

23)

(si

+
+

s*)

= 4p,
= 4?,

+ s0 =
2

(s z

s4)
ii)

+ *)*-(*, +
c3 , z 3 , s 4 ,

=4r;

from which we have

Si,

included in the formula

<

-r.

122 Algebraic Solution of

the Cubic

and Biquadratic.

We
and
a, j3, y,

now

also those of the reducing cubic (2), in

proceed to express the roots of Euler's cubic (1)> terms of the roots

S of the given biquadratic in x. Attending to the remarks above made with reference to the signs of the radicals, we

may

write the four values of z = ax + b as follows

aa +
0/3

b b

= -

ay +

= -

a$ + b =

from which

may

be immediately derived the following expres:

sions for p, q y r the roots of Euler's cubic

q-

7 +

-0-8)

s
,

(4)

Subtracting in pairs the equations (3), and making use of we the relations above written between p, q, r and 2, 3
t,

easily establish the following useful relations connecting the differences of the roots of the cubics (1) and (2) with the diffe-

rences of the roots of the biquadratic :-

4(q -r) 4(r

= 4*(0 Z -

ft)

p)
q)

= 4*(ft = 4
3

ft)
ft)

- - a'Q3 - - 2 (y

y) (a- S),
a)

8),

(5)

4(pFinally,
ft
.

(ft

= -

(a

- 0)(y -S).
of the
relation

from these equations, by aid

ft

ft

0,

we

derive the values of

ft, ft, ft

in terms of

P) 7>

S, viz.,

(6)

Examples.

123

EXAMPLES.
1.

Show

that the

two biquadratic equations

have the same reducing cubic. '2. Find the reducing cubic of the two biquadratic equations
x*

6lx~

Sx

+ w' + w 3 - Wmn +

(4wn
3

P)

0.
3
-I-

^s.
3.

3m0 - (m
.''

w3 )

0.

Prove that the eight roots of the equation

{x*

6lxz

-f

3 (4;>m

z ) }

64

3
(/

+ w3

-f

n3 - 3//)

are given

by the formula

i/l
(Compare Ex. 20,
4.

-f

m + n+ *l

-f

(am

+ orn + <v' +

or

w+

n.

p. 34.)

If the expression

V^

+ m

-f

-f

\/

+ wm +

ai-n

-t-

\/

-t-

urm

-f

ww

be a root of the equation

3T2 =
determine H,
/,

0,

/ in

terms of

/,

m, n.

A
5.

ii*.

11=

-I,

1=1 Imn, J = -

(w

3
)
.

cular cases

Write down the formulas expressing the root of a biquadratic in the when / = 0, and / = 0.

parti-

6. If the biquadratic has two equal roots, prove that the reducing cubic has equal roots, and conversely.
7. If the biquadratic has three roots equal, prove that 0, ,7=0. reducing cubic vanish, and consequently

two

all

the roots of the

/=

8.

If the biquadratic has

two

distinct pairs of equal roots, prove that

two of the

roots of Euler's cubic vanish,

9.

and consequently

G=

1 0, a

I2H"2

0.

Prove the following relations between the biquadratic and Euler's cubic with
:

respect to the nature of the roots

(1).

When
and

the roots of the biquadratic are all real, the roots of Euler's cubit-

are all real

(2).

positive.

the roots of the biquadratic are all imaginary, the roots of Euler's cubic are all real, two being negative and one positive. (3). When the biquadratic has two real and two imaginary roots, Euler's
cubic has

When

two imaginary roots and one real positive root. These results follow readily from equations (4) when the proper forms are sub/3,

stituted for o,

7, 8 in the values of/?, q, r.

Tt is to

be observed that

all possible

124

Algebraic Solution of the Cubic and Biquadratic.

cases are here comprised, the biquadratic being supposed not to have equal roots.
It follows that the converse of each of these propositions is true.

Hence,
all

if

Euler's

cubic has all

its roots real

;

and

positive,

we may

conclude that

the roots of the

biquadratic are real

if

roots of the biquadratic are imaginary

Euler's cubic has negative roots, we conclude that all the and if Euler's cubic has imaginary roots,
;

we

conclude that the biquadratic has two real and two imaginary roots.
10.

Prove that the roots of the biquadratic and the roots of the reducing cubic

are connected

by the following

relations

the roots of the biquadratic are either all real, or all imaginary, (1). the roots of the reducing cubic are all real ; and, conversely, when the roots of the reducing cubic are all real, the roots of the biquadratic are either all real or
all

When

imaginary.

(2). When the biquadratic has two real, and two imaginary roots, the reducand, conversely, when the reducing cubic has ing cubic has two imaginary roots imaginary roots, the biquadratic has two real and two imaginary roots.
;

the two cubics

11. If

These results follow readily from the preceding example, since the roots of (1) and (2) are connected by a real linear relation.

is positive,

the biquadratic must have imaginary roots.

roots of Euler's cubic cannot be all positive.
real

For in that case the

12. If

Jis negative, the biquadratic has two

and two imaginary

roots.

For the reducing cubic has in that case two imaginary

13. If

roots (Ex. 12, p. 33).

H and / are both positive,

/ is

all

the roots of the biquadratic are imaginary.

;

For, since

positive, the reducing cubic has a real negative root

there-

fore also Euler's cubic has a real negative root, since t = a 2 - H, and is positive ; and this is case (2) of Ex. 9. It is implied in this proof that the leading
coefficient

is

positive

if

a/ be substituted

for

,7 in

the statement of the proposition

no

restriction as to the sign of a is necessary.

14. Express, by the aid of the reducing cubic, /and of the roots a, 0, 7, 5. (See Exs. 16, 18, Art. 27.)

/in terms

of the differences

15.

in terms of

Express the product of the squares of the differences of the roots / and 7.
of the equations
:

o,

7, 8

By means

(5)

above given, and the equation

(2), p. 82,

we

ob-

tain the result as follows

16.

What

is

the quantity under the/naJ square root (viz., that which occurs

under the cube root in the solution of the reducing cubic) in the formula expressing a root ? Ans. 277 2 - 7 3
.

17.

Prove that the

coefficients of the equation of

4 3 o# + 4ia; +

squared differences of the

biquadratic equation

62#2 +

403*

04

may

be expressed in

terms

>

H>

I>

and

/.

Second Solution by Radicals of

Removing the second term from the
65"
4
equation,

the Biquadratic.
obtain

125

we

0;

and changing the signs of the

roots,

we have
4
2

/-3#

These transformations leave the functions (o - 0) 2 &c., unaltered; but G becomes - G, the other coefficients of the latter equation remaining unchanged therefore G can enter the coefficients of the equation of squared differences in even
;

powers only.
ducing a
,

And by
/, /.

aid of the identity of Art. 37, G~

may

be eliminated, intro-

H,

In a similar manner we
j8,

of the differences of the roots a,

7, 5

may

prove that every even function. be expressed in terms of o> H, I, J,

may

the function

of odd degree not entering.

,.62.

Second Solution by Radicals of the BiquaLet the biquadratic equation

ff#
4

dratic.

+ 4bz? +

6cx*

+ 4dx +

be put, as before, under the form

s
4

6J5Ts

4s + a / - 3ff
2

0,

where

= ax +

b.

We now
equation

assume as the general expression for a root of

tLis

a formula involving three independent radicals, y^p,

v/</,

</r.

Squaring
2

twice,

and reducing, we have

r),

(s

- qr - rp pqf = 4pqr(2z + p + q +
z

or
2 4 - 2 (qr + rp

+pq) z

- Spqrz +

(qr

+ rp +pq) z -4(p + q + r)pqr =

s,

0.

Comparing
easily find

this equation

with the former equation iu

we

G
whence
p, ^, r are the roots of the equation
+,

cfl

(12H*

a* I)

e - QHGt + G* =

0.

126

Algebraic Solution of the Cubic and Biquadratic.

readily transformed into Euler's cubic, the substitution directly

This equation
or

may be

making

and putting
reduce
it

for

its

value in terms of

H,

/,

and

J",

we may

to the standard

form of the reducing

<7=0.

cubic, viz.,

It

is

important to observe that in the present method of soto that of of z has, in

lution

we meet with no ambiguity corresponding

;

Art. 61

for the expression here

assumed as the value

virtue of the double signs of the radicals contained in it, only four values, while the form assumed for z in the preceding Article

has eight values.

This appears from the identical equation

^

(/P +

</<!

^rf - p - qnumber

r,

which shows that the number of

2

distinct values of the radical

expression of the present Article is the

same

as the

of

values of (\/p + \/q + \/r) , namely four. In order to express p, q, r in terms of the roots

a, ]3, 7, S of
/3,

the biquadratic,

we

have, giving to x the four values a,

7, 8,

The student may

these four.

easily satisfy himself that

no combination

of the signs of the radicals can lead to

any value different from

s 2 Ss

From

the values of s 2 +

s3

s,

s4 ,

and

3iSi,

we

obtain

Resolution of the Quartic into

its

Quadratic Factors. 127

From
tion

these

and similar equations we have, employing the rela= - 2pqr, the following modes of expressing p, q, r in
,

terms of the roots

/3,

7, S

)3y-aS

SO

ya-(l$ = -q = a^+b=
r

8G
-

__

= a

-+ b

63.

Resolution of the Uuartic into

Let the quartic

its

Uuadratio

Factors.

be supposed to be expressed as the difference of two squares* in the form

this expression,

and comparing it with Multiplying the given quartic by we have the following equations to determine M, N, andfl: z z JT- = b z - ac + a 2 0, =bc-ad + 2ab0, ac. =(c + 2a8)
,

N MN Eliminating M and N from these equations, we find

4a 3
3

(ac

2 - 4bd + 3r) aO + ace + 2bcd ad

eb

0,

which

is

From

the reducing cubic before obtained. this equation we have three values of B

with three corresponding values of

(0i,

2,

#a)>

z
,

JfJV, JV

8
;

and thus

all

the coefficients of the assumed form for the quartic are deter-

of the quartic to the difference of two squares was the method This mode of solution of the equation of the fourth degree. for the employed solution is due to Ferrari, although by some writers ascribed to Simpson (see Note A).

The reduction

first

The method explained

in the following Article, in

which the quartic

is

equated

directly to the product of

two quadratic

factors, is

due to Descartes.

128 Algebraic Solution of the Cubic and Biquadratic.

mined in three
distinct

that to each value of

M corresponds

ways; moreover,

it

should be noticed

a single value of

N,

since

MN = be
The
quartic
(ax-

- ad + 2abO.

+ 2bx + c + 2aO)* - (2Mx + N) z

may

plainly be resolved into the two quadratic factors

When

6 receives the three values

ft, ft, ft,

we

obtain the three

pairs of quadratic factors of the original quartic, is completely solved.

and the problem

solution

In order to make clear the connexion between the present and the solution by radicals, let us suppose that the
a,

roots of the quadratic factors in the order above written are

]3,

7 and

dratic factors

and that the roots of the remaining pairs of quaare similarly 7, a and )3, S a, /3 and 7, 8. We
;

have, therefore,

+ -(b a

),

j3

+ S=

), -?(6+Jf a
2

y+

$=-z

where
2,

Jf3 = ^/b - ac

Subtracting the last equations in pairs,

we

find

and

since

a +

/3

+ 7 + 8 =
= - Mi +

-4-,
z

we

obtain

aa +

M+

ay +
aS +

b
b

= = -

M+ M\

Resolution of the Quartic into

its

Quadratic Factors. 129

It appears, therefore, that the roots of the biquadratic are here expressed separately by formulas analogous to those of Art. 61.

The values

of

JP,

viz.

M*, MS,

J/" 3

2
,

are in fact identical with

There the roots of Euler's cuibo in the preceding Article. exists also with regard to the signs of the radicals involved in
MI,

M,

a restriction similar to that of Art. 61

since, in

virtue of the assumptions above

of the quadratic factors,

made with

respect to the roots

we have

the equation

which implies the following relation

(see

Ex. 20,

p. 52)

and by means
stricted in the

manner explained

of this relation the signs of J/i, 3 are rein the previous Article.

M^

By
all

aid of the equation last written

roots,

we can

eliminate

M*

from the expressions for the

and thus

obtain, as in Art. 61,

viz.,

the roots of the biquadratic in a single formula,

in

which the radicals MI = i/b*-ac + d*9i, and

M = </b z 2

are taken in complete generality.

EXAMPLES.
1
.

Form

the equation whose roots are A,

(3y

/t,

v, viz.,

a5,

70 +

5,

aj8

+ y5.
we have

Adding the

last coefficients of

the quadratic factors of the quartic,

70

4-78=
where
61, 0-, 03

40 3

2 -,

are the roots of the reducing cubic

hence the required equation.

3

Ans. (ax (Compare Exs.

4, 5, Art. 39.)

2c)

ll(ax

2c)

16

/=

0.

130

Algebraic Solution of the Cubic and Biquadratic.

2. Express, by means of the equations of the preceding example, the roots of the reducing cubic in terms of the roots of the biquadratic.

2c

Substituting for

its

value in terms of

a, 0, 7, S,

we

find immediately

1201

= 2A =
=
2/4

/x

(y
(a

a) (0

S)

(a

/3)

(7

S),

120 2
1203
(Cf. (6), Art. 61.)
3.

-As

0) (7

OB (7

7) (a
o) (0

8),

2v

- A-/i =

(/3-

7 )(a

8)

S).

Verify,

by means

of the expressions for

0i,

2,

03 in

Ex.

1,

the conclusions of

Ex. 10, Art. 61, with respect to the manner in which the roots of the biquadratic and reducing cubic are related.
4.

Form

the equation whose roots are the functions

(07-oS)G3 + 7-o-8),

7 - ,35) {7 + a-0-8),

J-(aj8

- 78)

(a

+ )8

-7- 5).

From

the quadratic factors of the quartic

4J*i

we

find

- =0+7-0-5,
M\N\ be

2^i -= 07-08;
2ab6\

also

- ad +

=-

a2 <j>i,

tho roots of the required cubic being represented "We obtain, therefore, the required equation

by <i, 02, <j>3by a linear transformation

b*I(a?<})

of the

reducing cubic.

Ans.
5.

(a~<f>

+bc70 -

ad)*.-

be

ad)

2b*J

0.

Form

the equation whose roots are

7 - 08

8
S'

o/3

- 78
the corresponding root

+
If
<f>

7-0-5'
we

7+

0-0-

+ 0-7-8'

denote any one of these functions indifferently, and

have, employing former results,

of the reducing cubic,

*~~M*~ ~&-ac~+~a* i
reducing cubic.

_ Mir_bc-ad+ 2abB
'

and thus we obtain the required equation by a homographic transformation of the This formula may be put under the more convenient form

by means

of

which we obtain the required cubic in the following form

(0 +

b)*

2 (a*I- 121T ) (*0 +

i)

6HG(a<t>

b)

0,

3 which, expanded and divided by o , becomes

2 (abe +
(Cf.

2b*d- Zacd] Q +

Pe- ad- = 0.

Ex.

14, p. 88.)

Resolution nf the Quartic into Quadratic Factors.

(>.

131

Fonn

the equation whose roots are

These are the three values of

before, one of these values

N*

in the foregoing Article.

Representing, as

by <, we find that the required equation may he obtained from the reducing cubic by means of the homographic transformation
2

bed- ad z -

c-aQ
7.

Form

the equation whose roots are

7 - aS
'

_ __ _
ya-jSS

7 + a)38-(3+8)7a'

(a

+ 0) 7 5 - (7 + 5) a/3

The required equation

transformation

is

obtained from the reducing cubic by the homographic

cd

^~
This result
cals,

be

may be derived from Ex. 5 by changing the roots into their reciproand making the corresponding changes in the coefficients.

64.

The Resolution of
Second Method.
ax*

Factors.

the Quartic into Quadratic Let the quartic

be supposed to be resolved into the quadratic factors

a (x 2 + 2px +
q) (x~

+ 2p'x +

q').

We have, by
7\

comparing these two forms, the equations

yj

y-7

/>

p+p'=2-,
I*

q+

q'+4pp'=6-, U
fifth

= pq' + p'q

2-,
U

M' = II

(1)

If

now we had any

equation of the form

could eliminate p, p', the several values of 0.

we

q, q';

and thus

find

an equation giving

f = 0, or q + <f equation might be assumed to be pp = ; and in each case would be determined by a cubic equasince each of these tion, functions, when expressed in terms of

The

fifth

<

K2

132

Algebraic Solution of the Cubic and Biquadratic.

It is

the roots of the biquadratic, has three values only.

more

convenient, however, to

assume
2c

the two functions of p,

p', q,

second of equations
equations,

(1).

We

q here involved being equal by the easily find, by the aid of those

4abc - 2a?d

8bd>

and eliminating p, p',

q, q',

by means

of the identical relation

(?

+p'W + f]

- (/>/ -p'vY + (PI + p'v'Y,

there results the equation

+
which
is

J=

0,

the reducing cubic obtained

pp', or q

by the previous methods

of solution.

Having thus found

q',

we may complete the

resolution of the quartic

by means

of the equations (1).

The reason
of the
it

for the

the form of the fifth

assumption above made with regard to equation is obvious. From a comparison

assumed values of
is

with the equations of Ex.

1,

Art. 63,
;

appears that

the same as 9 in the preceding Article

q, q',

therefore
to

we

foresee that the elimination of p, p',

and must lead

obtained.

identical with the reducing cubic before an equation in In general, if represent any function of the differ/u,

ences of A,

v,

ences of

a, |3,

and consequently an even function of the differ7, 8 (see Ex. 18, Art. 27), the equation whose
cannot involve any functions
r/,

roots are the different values of

of the coefficients except

H,

J,

and

J.

be assumed equal to any of the expressions in the^second of the following examples, the equation in whose roots are the
If
different values of this expression is

formed as in 4 the above in-

stance

by the elimination of p, p',

q, q'.

Transformation of Biquadratic into Reciprocal Form. 133

EXAMPLES.
1.

Resolve into quadratic factors

Comparing

this

form with the product

P+%
ure find the following equation for

+f
p
:

and putting
a~<(>

j?2

= pz +
,

H=
- lap

this equation,

when

divided

3 by a becomes

4a 3
2. If

3
<j>

+J= 0.

a quartic be resolved into the two quadratic factors

is determined by a cubic equation sponding to each of the following types

prove that

<f>

when

it

has all possible values corre-

q-q

pq'-p'q

pq'

-p'q

and by an equation of the sixth degree when

p> 9t

it

has

all

values corresponding to

P - P'J

- 9',

PQ' -p'q,

p2 - *?

Expressing these functions in terms of the roots, the number of possible values of ach function becomes apparent.

65. Transformation of the Biquadratic into the Reciprocal Form. To effect this transformation we make

the linear substitution x = ky + p in the equation

ax*

+ 4bx? + 6cx* + 4dx +

0,

which then assumes the form

+ V* =
where
Ui = ap +
b,

0,

= ap* + 2bp +

c,

U = ap
3

+ 3bp z + 3cp +

d,

&c.

(Cf. Art. 35.)

If this equation be reciprocal,

we have two equa-

tions to determine k

and

p, viz.,

ak*=

tfZTA-ZTi

134

Algebraic Solution of the Cubic and Biquadratic.

&,

eliminating

we have

the following equation for p

aU^ - Cr Cr4 = 0;
2
i

and

since
,

_ =

IT*

~U\

2 3 _ ap + 3bp + 3cp + d = ~~' ap + b

there are two values of

k,

equal with opposite signs, correspond

ing to each value of p.

The equation

aV?when reduced by

Z7i Z7i

0,

the substitutions (Arts. 36, 37)

becomes

20C?
which
is

+ (a*I- 12H*)

C? - SGHU, -

G* =

0,
if

(1)

a cubic equation determining Ui = ap'+b; and

we put

is

determined by the standard reducing cubic

4a*0*-IaO +
This transformation*
dratic
;

J=

0.

may

and

it is

important to observe that the cubic (1)

be employed to solve the biquawhich

here presents itself differs from the cubic of Art. 62 only in having roots with contrary signs.

We proceed now to express k and p in terms of a,

roots of the biquadratic equation.

]3,

7, S, the

obtained

by putting x = ky +
,

p, is

Since the equation in y y reciprocal, its roots are of the

form y l9 yz ,
y2

hence we

may write

y\

* This method of solving the biquadratic by transforming

it

to the reciprocal
vol.
i.

form was given by Mr. S. S. Greatheed in the Camb. Math. Journ.,

Transformation of Biquadratic into Reciprocal Form. 135

and, therefore,
(a

p)($ -p)

(/3

p)(y

p)

= k\

from which we find

P =
|3

+ 7 - a

-~8'

and

2
A:

-4

An important geometrical interpretation may be given to the quantities k and p which enter into this transformation. Let the distances OA, OB, 00, OD, of four points A, B, C, D,
on a right line from a fixed origin

on the

line be

determined

by

the roots

a, j3, 7, S,

of the equation
e

ax^

+ 4 bx* + 6 ex2 + 4dx +

be the centres
;

=
;

also let

1?

0o,

and

l9

F{\

F F
Zy

'
;

F F
3,

3',

the foci of the three systems of involution determined three following pairs of quadratics
:

by the

\ f

s\

f\

We have then the equations

O^B 0,0 = 0,A
.
.

OtD =

8
,

OiJF !

&c.,

which, transformed and compared with the equations

O - p)(y - P)
since

p)(8

ri

8
A;
,

&c.,
,

prove that the three values of p are 00i, 002 , 003 the distances of the three centres of involution from the fixed origin 0. Also

OiF* =

k~,

k has six values represented geometrically

by

the distances

where 0i^i + 0\F{ =

opposite directions.

0, &c., as

the distances are measured in

136

Algebraic Solution of the Cubic and Biquadratic.

We can

tions of the centres

from geometrical considerations alone find the and foci of involution in terms of a, /3,
results just established, as follows
:

posi7, S,

and thus confirm the

Since the systems

{F^BFiC] and (FjAFfD}

+

1111
F,C~ FtA
+
1 1
1
~r

are harmonic,

FtD'

and
0,

if

we

x represent the distance of FI or have

1

FS from

the fixed origin

X-

/3

-j

x-a

f*.

Solving this equation,

we

find

or

=--

whence

p=

-,

k,

+ OF:
~
,

k=

--

OF, - OF; =

EXAMPLE.
-.

Transform the cubic

to the reciprocal form.

The assumption x = ky +

p leads to the equation

+
The values

3J3" 2 Z7i 3

0,

where

Z7i

ap

b.

of p are easily seen to be

07 ~
P+7-

ya-P
7 + a - 20'
a

a/3-7 2

2a'

/8

27'

The

taken on the axis such that A'

geometrical interpretation in this case is, that if three points A', B' C' be is the harmonic with respect to conjugate of
,

and

'

C,

of

B with respect

to

C and A, and
:

C' of

C with

respect to -4

and

then

we have

the following values of p and k

p=

OA +
*

OA'
'

* =

OA-OA'

"2

'

For the values of OA', OB', OC', in terms of

a, 0, 7, see

Ex.

13, p. 88.

Solution of the Biquadratic by Symmetric Functions.

66.

137
'"

Solution of the Biquadratic by Symmetric Fuuctions of the Roots. The possibility of reducing the solution of the biquadratic to that of a cubic by the present method depends on the possibility of forming functions of the four roots which admit of only three values when these roots are /3, y, S, interchanged in every way. It will be seen on referring to Ex. 2,
>

Art. 64, that several functions of this nature exist. These, like the analogous functions of Art. 59, possess an important property to be proved hereafter, viz., any two -such sets of three are
so related that

any one function of either set is connected with some one function of the other set by a rational homographic

relation in terms of the coefficients.

of the present solution we employ the functions already referred to in Art. 55, since they lead in the most direct manner to the expressions for the roots of the bi-

For the purposes

quadratic in terms of the coefficients. proceed accordingly to form the equation whose roots are the three values of a + 0/3 +
2

We
3

y +

SV

when the

roots are interchanged in every

= way, and 9

1.

These values are

and

since

S (a - |3)
we

77"
2

3Sa 2 - 2A - 2/x - 2v = - 48
t2 ,

find the following values of t^

tz

2A- M -v

2fjL-v-\

H
.

2v-\-

whence

^ +

t2

tz

= - 3
a

138

Algebraic Solution of the Cubic and Biquadratic.

Again, since

and

we have

also

Hence the equation whose

roots are t^
2

2,

4 becomes
-

~-<h

or, substituting for (r

its
3

value from Art. 37,

(^

J2")

- a2 /(^ + JT) +

J"=

0,

which

is

substitution

transformed into the standard reducing cubic by the a?0. aH +

H=

To determine

a, |3, 7, 8

we have

the following equations

,

a + ]3

-7- S = 4

along with

a +

/3

+ y + S =

from which we find

/3

Solution of Biquadratic by Symmetric Functions.

139
the

We have also from

equation

the above values of

yij

\/^2,

A/%

by means

which one radical can be expressed in terms of the other two, and the general formula for a root shown to be the same as those previously given.
of

It

is

to give

some account

convenient, in connexion with the subject of this Article, of two functions of the roots of the biqua-

which possess properties analogous to those established in Art. 59 for corresponding functions of the roots of a cubic.
dratic

Adopting a notation similar

to that of the Article referred to,

/u,

we may
ing form

write these functions in terms of A,

:

v in the follow-

aS)

a>

(ya

+ /3S) + W 2

(a/3

+ 78),

By means
the form

of the equations of

Ex.

1,

Art. 63, these functions

can be expressed in terms of the roots of the reducing cubic in

JZ

01

+ W02 + 0> 2 0,,

M=

0!

-I-

W2

W0 8

They may
as follows
:

also

sent Article connecting

be expressed, by aid of the equation of the pret and 0, in terms of the values of t it t 2 t 3 ,
,

The
cubic.

functions

and

M are

as important in the theory of

the biquadratic as the functions of Art. 59 in the theory of the

The cubes

of these expressions are the simplest functions

of four variables

which have but two values when the variables

;

are interchanged in every way they are the roots of the reof the ducing quadratic reducing cubic above written, and underlie every solution of the biquadratic which has been given.

140

Algebraic Solution of the Cubic and Biquadratic.

EXAMPLES.
1
.

Show

that

and

M are functions of the

by
h,

differences of a,

7, S.

Increasing
2.

o, 0, 7, 5

L and

JIf

remain unaltered, since

1 -f

+ w2 =

0.

To

find in terms of the coefficients the product of the squares of the difiera,
,

ences of the roots

7, S.

From

the values of

and

M in terms of
2

0i, 0o, #3,

we
(

find easily

120i=
120 2
12 03

L+
a>

M,

L- M=
Z - wJf =
wZ - arJf =

08

- y)

8)

(o>

),

=
=

Z + w.,
+ a>W,

(7
(a

a) (0

8) (a>
2

),

a>i

- /B)

(7

8)

(a>

).

Again, from these equations, multiplying the terms on both

sides together,

and

remembering that

0i, 02> #3

are the roots of

we

find

also,

adding the squares of the same terms,

we have
2
)

1LM = 24 ~ = OB - 7) 2 (a - S) 2 + (7 and, since

{ft

S)

- &)*

(7

- 8) 8

(Z

- Jf 3 ) 2 =

(Z

+ Jf3

1
)

- 4Z 3 Jf 3

substituting for these quantities their values derived from former equations,
finally

we have

3. Show by a comparison of the equations of the present Article and Art. 59 that the results of the previous Article may be extended to the biquadratic by changing

0-7, 7-0, a-

into

-OB -7) (a -5), - 7 -a)()3 - S), (

(a

(7- S),

respectively; and, consequently,

H into

J,

and

into 16/.

67.

Equation of Squared Differences of a Biqua-

dratic.

In a previous chapter (Art. 44) an account was given of the general problem of the formation of the equation of differences. It was proposed by Lagrange to employ this equation in practice for the purpose of separating the roots of a

given numerical equation

and with a view to such application

Equation of Squared Differences of a Biquadratic. 141

he calculated the general forms of the equation of squared
ferences in the cases of equations of the fourth
dif-

and

fifth

degrees

wanting the second term (see Traite de la Resolution des Equations Ntimeriques, 3rd ed., Ch. v., and Note in.). Although for
practical purposes the methods of separation of the roots to be hereafter explained are to be preferred ; yet, in connexion with

the subjects of the present Chapter, the equation of squared differences of the biquadratic is of sufficient interest to be given

proceed accordingly to calculate this equation for a It will appear, biquadratic written in the most general form. in accordance with what was proved in Ex. 17, Art. 61, that
here.

We

the coefficients of the resulting equation can all be expressed in

terms of

#,

H,
is

I,

and

J.

The problem

equivalent to expressing the] following product in terms of the

coefficients of the biquadratic

The most convenient mode of procedure is to group these six factors in pairs, and to express the three products (which we denote by Hi, Eb, Ila) separately in terms of the roots of the reducing cubic, and finally to express the product HI IIj Ila in
terms of
a,

H,

7, J.

and,

by

aid of the results of Art. 61

we

easily derive the following expressions for

hence, without difficulty,

T

16-J (~pf\
801

<j>

+ 4--

Introducing

now

for brevity the notation

lQH=a-P,
Hi becomes

4/=

Q,

167= a?R,
the result of
(8Jty
3

+ 80i$ - 480 2 03-

Reducing the product Hi

3

n 2 n 3 by

Example

18,

page 89,
2
)

we

obtain

+ SQ^ - (4Qcj> 2 + 18Jty)2

+ 12QV + 36Q2ty + 27.K = 0.

we have the equation of squared differences exFinally, restoring the value of pressed in terms of P, Q, J2, as follows
:

6
4>

+ 3 J>5 + (3P2 + 2 Q) + (ps + 8P Q _ 26K) ^ + (6P 2 Q - 7Q2 - 18P22) f + 9Q (PQ - 6JZ)^ + 4Q3 - 27-fl 2 = 0.
<j>4

142

Algebraic Solution of the Cubic and Biquadratic.

also in terms of a,

"We give for convenience of reference the result

H,

I,

J*

16(3847/

/- 7a2 / 2 - 2S8a#/)<J> 2 +

1152

3 2 (2HI- 3a/)/^ + 256 (J - 27/ )=0.

It should be observed that the value

above obtained for

ITi

can be expressed as a
,

quadratic function of 0i by aid of the equation Oz 0$

calculation

9\~
Q\

and the subsequent

this quadratic

might have been conducted by eliminating

between

and

the reducing cubic.

68.

Criterion of the Nature of the Roots of the

Before proceeding with this investigation
it is

Biquadratic.

necessary to repeat

what was before

stated (Art. 43), that

when

any condition with

respect to the nature of the roots of an algebraic equation is expressed by the sign of a function of the coefficients, these coefficients are supposed to represent real

numerical quantities. It is assumed also, as in the Article referred to, that the leading coefficient does not vanish.

Using

as before

to represent that function of the coef-

ficients (called the

discriminant] which,

when

multiplied

by a

positive numerical factor, is equal to the product of the squares of the differences of the roots, we have, from the results estab-

lished in preceding Articles, the equation

6

03

7 ) (7

0)'(

where

P - 27 J
of

z
.

It will be found convenient in discussion

of the
(1)

what follows
roots
is

nature

the

to arrange the under three heads,

is

according as
(1)

vanishes, or (2)

negative, or (3)

positive.

When A vanishes,

the equation has equal roots.

This is evident

from the value of A above written. Four distinct cases may be noticed (a) when two roots only are equal, in which case /and J do not vanish separately (j3) when three roots are equal, in which case /= 0, and J = 0, separately (see Ex. 7, Art. 61) (y) n-hcn
;
;

* The equation of squared differences was first given in this form by Mr. M. Roberts in the Nourelles Annales de Mathematiques, vol. xvi.

Criterion of
tiro distinct

Nature of Roots of Biquadratic.

143
the

pairs of roots are equal, in which case

we have

conditions

0, a~

I-12H*

hence these two equations, = 0, are equivalent to two indepenwith the A equation along dent conditions only. Finally, we may have (S) all the roots
;

readily proved by means conditions imply the equation

(Ex. 8, Art. 61). It can be of the identity of Art. 37 that these

A =

equal; in which case

may

independent conditions be written in a form analogous to the corresponding conditions

in case (4) of Art. 43.
(2)

be derived from Art. 61 the three = 0, I = 0, and J = 0. These may

When A
for

is

negative, the equation has two real

and two ima-

ginary roots.
roots
;

This follows from the value of

all

in terms of the
;

when

the roots are real

is

plainly positive
//

and

the proper imaginary forms, viz. h k ^/ - 1, are substituted for a, /3, y, S, it readily appears that also when all the roots are imaginary.

when

k'^/-l,
is

positive

(3)

WJien

is

positive, the roots

real or all imaginary.

of the equation are either all This follows also from the value of A, for

we can show by substituting for a, |3 the forms h k */ - 1 that A is negative when two roots are real and two imaginary. In the case, therefore, when A is positive, this function of the
coefficients is not

by itself

sufficient to
it

nature of the roots, for

roots

remains

still

determine completely the doubtful whether the

are

all real

necessary to discriminate

or all imaginary. The further conditions between these two cases may, however,
:

be obtained from Euler's cubic (Art. 61) as follows

In order

that the roots of this cubic should be all real and positive, it is necessary that the signs should be alternately positive and

negative and when the signs are of this nature the cubic cannot have a real negative root. can, therefore, derive, by the aid of Ex. 9, Art. 61, the following general conclusion appli;

We

cable to this case

is positive the roots of the biquadratic are all imaginary in every case except when the following conditions - 12H* negative ; in which are fulfilled, viz. negative, and a*I
:

WTien

case the roots are all real.

144

Algebraic Solution of the Cubic and Biquadratic.

EXAMPLES.

1.

Show

that if j? be positive, or if

H=

(and

not

0),

the cuhic will have-

a pair of imaginary roots. 2. Show that if J2"be negative, the cuhic will have unequal,
(2) (2)

its roots

(1) all real

is

and

two equal, or

(3)

two imaginary, according

as

Gz

(1) less than,,

equal
3.

to, or (3) greater

than

4jB" 3 .

If the cuhic equation

have two roots equal to o

prove

where
4.
a$a<i

= Hi

Fi
a\
z

^'
a\ as

If,

o3
ax*

i2 =
3bx
z

2J?i,

a%

If

3cx

+ d + k (x

- r) 3
cz )

he a perfect cuhe, prove

(ac
5.

b z ) r*

(ad

-bc)r +

(bd

0.

Find the condition that the cuhic

ax3

3bxz

Zcx

+d

may

be capable of being written under the form

I

(x

ai)

+ m(x-

3
i)

+n(x- 7 i) 3

where

ai,

j8i,

71 are the roots of the cubic

Comparing the forms, we have

a

b
c

= I + m + n, = lai + m(3i
=7oi 2 + w/3i
lai*
2

-d=
Also
a\ oi
3

3#i ai

+ 3d ai +

di

0,

&c.

Whence, multiplying these equations by

d\, 3ci, 3bi, ai, respectively,

and

adding,.

we

find the required condition

(adi
6.

aid)

(bc\

bic)

0.

If a,

ft,

y be the roots of the cubic equation

o# 3

3ai #

+
-

Sazx

+ #3 =

rationalize the equation

\x - a +
and express the

f3

x --y = 0;
o,
i,

result in terms of the coefficients

02,

3.

Ans. 125 Ui*

+ 3607/i+ 128^^ - 48.H" 2 =

0.

Examples.
7.

145

If 01,

/5i,

and

02, j3 2

be the roots of the quadratic equations

find the equation

whose
b\

roots are the four values of oia2 .

,

Let

J?i

0i c\

-H"2

c 2 c2

fa~.
(ci\(i?,<$P

Ans.

2$i3 2 <^>

2
i

eg)

4.ZZi _H"2

2
(|>

0.

N.B.

This and the two following Examples

may

be solved by expressing $ by

radicals involving the coefficients.

8.

Employing the notation of Ex.

.

7,

form the equation whose roots are the four

01

values of

+ 02
.

Let
Ans. (2i a z
2
tf>

(ai fa
is

a 2 fo)

<f>

-ffi 2 )

-Hi JT3

0.

In
9.

this

Example the
case,

resulting biquadratic
if
<J>

such that

G=

0.

In the same
<f>.

(ai

a2 ) 2, form the equation whose roots are the

several values of

Let

10. Show that when the biquadratic has a double root, the cubic whose roots are the values of p (Art. 65) has the same double root ; and find what this cubic becomes when the biquadratic has three roots equal. 11. If

H and J are

both positive, prove directly (without the aid of Euler's

cubic) that the roots of the biquadratic are all imaginary. in terms of the roots (Ex. 19, p. 52) that It appears from the expression for

k V 1. positive there must be at least one pair of imaginary roots h Noir diminishing all the roots by h, and dividing them by k (which transformations
ir hen
is

will not alter the character of the other pair of roots 7, S, nor the signs of

H and /),

the biquadratic

may be

put under the form

or

#*

+ 4px3 +
z
,

6cz*

+ px +
q

q,

where 6c = q
3c 2 ,
c*

whence

H=c-p
/=
qc +

I=

- 4p2 +
1)

2p*c-p*(q +

c (q

lp>

-#

and therefore
c

proving that y and 5 are imaginary

if

H and / are both positive

L

(cf .

Ex.

13, p. 124).

146

Algebraic Solution of the Cubic and Biquadratic.

two
distinct pairs of equal roots, prove directly the

12. If the biquadratic has

relations

=
In

12-ff2 ,

this case the biquadratic divided

by a assumes the form

where

= aQ x + ai,

k and
,

= a-/5

whence, comparing the forms

and

z4

6S"z2
2
,

+ Gz +

-Z"

3.H" 2 ,

we

find

3H=-

A;

0,

/-

3.H" 2

A*,

from which the above

relations immediately follow.

The student will

8,

easily estabit

lish the identity of these relations

with those of Ex.

Art. 61.

Also

should be

noticed that in this case only one square root is involved in the solution of the a) (x biquadratic (coming from the solution of the quadratic (x ) ).
13.

Find the condition that the biquadratic may be capable of being put under

the form

2px +
In
this case the second

q)

n.

and fourth

coefficients are

removed by the same trans-

formation, and the general solution involves only two square roots.

Am.
14.

0.

Prove that

7 vanishes

for the biquadratic

m(x -

w)

n(x
/3,

4.

15. If the roots of a biquadratic, a,

7, 5 represent the distances of four

from an origin on a right line prove that when these points form a harmonic division on the line the roots of Euler's cubic are in arithmetic progression, and the
points
;

roots of the cubic of Art. 62 in

harmonic progression.

16.

Form

points in a right line determined

the equation whose roots are the six anharmonic functions of four by the equation

=
The
six

0.

anharmonic

ratios are
1
4>1>
>

1
,

02,

03,

01

02

03

Examples.
* ~
(a

147
0i

0) (7
)

(7
Q3

8)
)

(13

A_ ~ - /* _ A

02

- 7) (a 0) (7 (o

S)
5)

_ _ 02^03 ~ M -y_ ~ 02 - 0i' p-\

~ (7
also the equation

g)

(j3

S) 5)

_7

) (

v= ~ v

A.

03-01
03

'

02

whose roots are

03
is

-7)

(a

-5),

7 -a) (0 -5),

(a

jB)

(7

5)

one of the cubics

16

v// 3 -

27/2

0.

The equation whose

of these cubics is

roots are the ratios,

with sign changed, of the roots of either

2

4A (0 2 where

I)

27 1 3

(0
.

I)

(see

Ex.

15, p. 88),

A = J3 _ 2 7/ 2
<

The roots of the equation in are the six anhannonic ratios. This equation can be written in a more expressive form, as will appear from the following propositions
(a).
:

The
:

six

anharmonic
1

ratios

may

be expressed in terms of any one of them,

'

as follows

1</,'

f-1

* '+ - *' +
s

From

the identical equation

(0

7) (a

S)

(7

a) 08

*)

(a

/B)

(7

8)

we have

the relations
0!

+
03

1,

02

+
0!

1,

03

02

1,

which determine
(b).

all

If

two
2
,

of the anharmonic ratios

the anharmonic ratios in terms of any one of them. become equal, the six values of
;

are

u and

each occurring three times and in this case I = 0. For suppose 0i = 02 we have then from the second of the above relations
;

0i

0i

+ 1 = 0,
or in

whence
and

0i

a>,

w2
(a),

substituting either of these values for

we find

all

the anharmonic ratios.

Also, since

- + A7^ r-- =
A.

0,

or2fc-*)=0,
3aj
2

A*

we have

/=

4i 03 +

0.

L2

148

Algebraic Solution of the Cubic and Biquadratic.

If one of the ratios
is

(c).

harmonic, the six values of

are

1,2,-,

each

occurring twice

and in

this case

7=0;
=v

for if

4>i

=-

~^
1,
A.

1,

or

2A -

fj.

0,

one of the factors of /(see Ex.

(d}.

18, p. 52).

These

results, as well as the

converse propositions,
:

may

"he

proved by

writing the sextic in

under the following form

_
17. Solve the equation

2)

(4,

18.

),

where

0*

1.

Express 2(

(7

5)

as a rational function of 0i,

2,

63

and ultimately

in terms of the coefficients of the quartic.

Am. 19.

128 2(02

2
3)

0i

~=
2

Express
(0
2

- 72

2
)

(a

- S2

2
)

+ 72 (

2 2
)

2
(/8

S2)2

(a

- ^(y* -

52 ) 2

as a rational function of

0i, 02, 0sis

This symmetric function

equivalent to
2
)

(fj?

- ^) 2 + (^ - A2

H-

(A

2 2 ,u
)

= 256 2 (0 3 -

2 3)

- -} !
(0i

20.

Form

the equation whose roots are the several products in pairs of the roots
the product of three factors of the type

of a biquadratic.

The required equation

is

Ans. (a<f21.
a,
j8,

2c<j>

3 *)

2 4/<^)
(fl(|>

2e(>

+e+ 16J<>

=
-,

0.

Form

the equation whose roots are the several values of

where

7, 8 are the roots of a biquadratic.

The

required equation

is

the product of three factors of the type

Ans. 4 (a<p

1bq>

+ cf-

I(a<^

23^ +

c)

+ J=

0.

Examples
22. Prove
1

149

9J

From

the expressions for a,

i

7, 8 in terms of

Oi, 6 2) 63,

we have

(a 2 ei+2s:
(

(a-p)*~~2tf
which may be expressed
23

(0 2

-0 3

2
)

a* 62
(03

+ 23:
2

-0i)

in terms of a, IT, I, J, as above.

if

J=
24.

0,

and

of the form 3p or 3p

l,p being a positive

integer.

Prove that

can be resolved into the sum or difference of two squares

if

J=aee +
Here
and
is

2bcd

ad*

eb*

c3

0.

aU

(ax + by + czf

(ac

b~) y

+ 2 (ad - be} yz+(ae- c2

+
(ae

z2 ,

(ac
if

&) y*+2 (ad

be) yz

- c2

z2

a perfect square

or

7=0.
25. If
a,
/3,

(ac-V}(ae-c*) = (ad-bcy,
7, 5 be the roots of the equation

=
solve, in

0,

terms of the coefficients

o,

i,

&c., the equation

\/a?

+ V^ -

)8

+ V^^ - 7 + X/2 -8 =

0.

When
is rationalized,

v" + A/is"+ \/7 + \/s"=

and the
coefficients substituted for a,
/3,

7, 5,

we have

(3o2-2i2) 2 =
Now,
substituting
ZTo?
iTi,

4.

Vz,

Us, Ui for

o,

i,

2,

3,

4,

and reducing, we

find

150
26.

Algebraic Solution of the Cubic and Biquadratic.

To
express the solution of the biquadratic in terms of a single root of the

reducing cubic.
Substituting

x1 +

p for x in the equation

ax*

4bx*

6cx*

+ 4dx +

0,

we have =
As there are here two independent make the assumptions
4

0.

variables at our disposal,

it is

allowable

to-

0,

Vix' 2

+ Us =

0.

Eliminating

x' z

and reducing

as in Art. 65,

we have

whence

J7a

0,

where

6 is a root of the reducing cubic,

and therefore

Ui

= ap +

= *a*6- H.

Again,

whence,

finally, since

a;

a/

p,

or ##

Z7i

ax*,

we have

J-^e-ZH -- ._
-

an expression which has only four values.

This expression might of course be obtained from the resulting formula of Art. 61
or Art. 63.
,.

The method of

arriving at

it

in the present

Example is a distinct method

of solving the biquadratic. 27. Prove that every rational algebraic function of a root 6 of a given cubic equation can in general be reduced to the form

Co

Ci6

Do + Did'
(h (ff\

Let the given function be

tions of 6 of

^-j-l >

where $ (0} and

fy(0)

are rational integral func-

any

order.

By

successive substitutions

from the given cubic each of

is

these

may be

reduced to a quadratic.

Hence the given function

reducible to the

form

Equating

this to the

form written above, and reducing by the given cubic, we ob-

tain an identical equation, viz.,

where

Zi, Iz are linear functions of Co, C\ t DO, D\.

"We have, therefore, the

three equations LQ

0,

L\

0,

L%

= 0,

to determine the ratios of Co, Ci, DO, -Oi.

Examples.

151

28. Prove that the solution of the biquadratic does not involve the extraction

of a cube root

when any

relation

among the

roots o,

)8,

5 exists

which can be ex-

pressed

by

Any

the vanishing of a rational function of a root 6 of the reducing cubic. rational function of 9 can always be depressed to the second degree, as in

the preceding example.

tion of a cube root
;

Hence the determination

of 6 will not involve the extracfor the

and the formula of Ex. 26 shows that the expression

root of the biquadratic will not then involve

29.

any cube

root.

Find the

relation

which connects the

3

roots of the biquadratic

when

the

equation

4p
is satisfied

-7p +

/=0
:

by each

of the following values of p

Ans.

(1)

0+ 7 - a _5 = 0,
(8)

(2)

+ 7 = 0,

(3)

(7-

a) (j8- 5)

- (a-) (7-8) = 0,
(6), (7)

(4),

7 -a5=0,

(5)

7 -a) 03-5)-(a-0)( 7 -8)=0,

/3- 7

= 0.

30. Prove the identity

2 (J3- 277 ) =

(a

*I- 3J? 2
:

K J2

This

may be
it

proved as follows

Putting a\

in the values of

1 and

/,

and ex-

panding, the form

readily appears that the part of

independent of a\

may be thrown into

Now,
ties

replacing o 2 , s, the values of Art. 37,

by Az,

3,

At, and substituting for the

latter quanti-

we

obtain the result.

Mr. M. EGBERTS.

31.

When

equal roots whose

a biquadratic has two equal roots, prove that Euler's cubic has two common value is

21

'

and hence show that the remaining two roots of the biquadratic in equal, or imaginary, according as 2J2T- 3a/is negative, zero, or
32.
last

this case are real, positive.

Prove that when a biquadratic has

two terms

(1) two distinct pairs of equal roots the of the equation of squared differences (Art. 67) vanish, giving the
;

conditions

A = 0, 2HI - 3aJ = and when it has (2) three roots equal, the last three terms of this equation vanish, giving the conditions I = 0, / = and show ; the equivalence of the conditions in the former case with those already obtained in
Ex.
8,

Art. 61,

and Ex.

12, p. 146.

ferences reduces in the former case to

Prove also that the equation of squared dif2 2 (0 + 12.B") 4 and in the flatter case to
<j>

(j>

CHAPTER VII
PROPERTIES OF THE DERIVED FUNCTIONS.
69.

Graphic Representation of tbe Derived

Let

tion.

APB

be the

curve representing the pothe lynomial f(%)) and

point on it corresponding to any value of the variable x =

OM. proceed to determine the mode of representing the value of f(x] Take a se-

We

at the point P.

cond point

on the curve,
Fig. 5.

corresponding to a value of x which exceeds ON. by a small quantity

h.

Thus

also

PM =/(#),
f(x +

QN=f(x

h).

The expansion
h)

of Art. 6 gives

=f(x)

+f(

or

But

Now, when A is indefinitely

and ultimately coincides with,

diminished, the point the chord ;

PQ becomes

approaches, the

Maxima and Minima

Values.

153

the angle becomes PTM. to the curve at tangent Also all terms of the right-hand member of equation (1) except the first diminish indefinitely, and ultimately vanish when h = 0.
;

PT

PEN

The equation

(1)

becomes therefore
tan

Prif /(*);
the value

from which we conclude that

function

(x)

assumed by the derived on the substitution of any value of x is represented by

the tangent of the angle

made with

the axis

OX by

the tangent at

the corresponding point to the curve representing the function f(x).

70.

Maxima and Minima Values of a Polynomial.

Any
value of x which renders f(x) a

Theorem.

maximum

or

minimum is a root of the derived equation f'(x) 0. Let a be a value of x which renders f(x] a minimum. We proceed to prove that /"(a) = 0. Let h represent a small increment or decrement of x. We have, since /(a) is a minimum,
/(a) </(a +
h), also

/(a)

</(a

h)

hence /(a + h) -/(a), and /(a -h} -/(a) are both positive, the following two expressions are positive
:

i.

e.

is very small, we know (Art. 5) that the signs of these expressions are the same as the signs of their first terms ; hence, in order that both should be positive,/' (a) must vanish ;

Now, when h

An exactly similar proof and, moreover, /"(a) must be positive. f shows that when /(a) is a maximum f (a) = 0, andf'(a} is negative. Thus, in order to find the maximum and minimum values
of a polynomial/(#), we must solve the equation /'(#) = 0, and substitute the roots in /(a:). Each root will furnish a maximum or

minimum

value, the criterion to decide between these being the

sign of /"(a?)
value

when

the root
is

is

substituted in

it

whenf"(x)
is positive,

is

negative, the value

is

maximum ; and when f" (x)

the

a minimum.

154

Properties of the Derived Functions.

The theorem of
ticle follows at

this

Ar-

once from
;

the construction of Art. 69

for
it is

plain that
is

when

the

value of /(a?)
as at

maximum,
6),

P, P' (Fig.

or a mi-

nimum,
parallel

as at p, p', the tan-

gent to the curve will be

to

the axis

OX,
Fig. 6.

and, consequently,

Fig. 6 represents a polynomial of the 5th degree. Correspond= ing to the four roots of /'(#) (supposed all real in this case), f viz. OM, Om, Om', there are two maxima values, HP,

OM

M'P', and two minima values, mp,

m'p', of the function.

EXAMPLES.
1.

Find the max. or min. value of

f(x)

2*2
/

4.

6.

/
x
(See
fig. 2,

= 4*+l,

= =

;; M ' I

- makes

Jr\yj

49
-,

a minimum.

p. 15.)

2.

Find the max. and min. values of

f(x)

36# + 14. = 6(*2-*-6), f'(x] = 6 (2*-l). x = 2 makes f(x) = 68, a maximum. 2# 3

3

x
3.

makes f(x] = -

67, a

minimum.

Find the max. and min. values of

Here
fix)

/ (^) =
7

has only one real root, x

and

it

gives a

minimum

value,

=-

345.

4.

Find the max. and min. values of

The

roots of f'(x) are, approximately,

value, the latter a

-0302,

1-1031.

The former

gives a

maximum

minimum.

(See

fig. 3, p. 16.)

Constitution of the Derived Functions.

71. Rolle's

155

Theorem.

Between two consecutive real roots

there lies at least one real root

a and b of the equation f(x) =

the

of

= 0. equation f(x) For as x increases from a to

b,

/(#)> varying continuously

by increasing and then diminish, It must, or must begin by diminishing and then increase.
from,

f (a)

to/(5), must begin

therefore, pass

through at

least

one

maximum

or

minimum

This value (/(a), value during the passage from f(a) to f(b). suppose) corresponds to some value a of x between a and b,

which by the Theorem of Art. 70

is

a root of the equation

The

We

figure in the preceding Article illustrates this theorem. and observe that between the two points of section

there are three

maximum

or

minimum

values,

and between the

two points B and C there is one such value. It appears also from the figure that the number of such values between two
consecutive points of section of the axis
is

always odd.

Corollary.

Two

consecutive roots of the derived equation

may
and

not comprise between them any root of the original equation, never can comprise more than one.

The first part of this proposition merely asserts that between two adjacent zero values of a polynomial there may be several maxima and minima values and the second part follows at once from the above theorem for if two consecutive roots of f(x] = comprised between them more than one root of f(x) = 0, we
; ;

should then have two consecutive roots of this latter equation

comprising between them no root of f(x] = dictory to the theorem.

72.
roots of the equation f(x)

0,

which

is

contra-

Constitution of the Derived Functions. Let the = be cii, a 2 a 3 an We have

,

f(x) = (x-

<n) (x

a z )(x

a s)

(x

a n ).

In

this identical equation substitute

y + x for x

156
where
q\

Properties of the Derived Functions.

= x=
(x

cti

+ x - az + x - a3 + - az + )
(x

+ x - an,
.
.

q*

ai) (x

-o

x)

(x

- a s) +

-f

(x

a^i)(at

- aw ),

(#-) (# -a
(x

3)

(a?

-a n ) +
+

(a:

a^(x
ai) (x

-a 3)
- az )

(x
.

a) +
a_i),

(x

(x

d) (a?

- a2 - as ) (a? )

(*-

a n ).

We have,

again,

fly + x)- f(x]

+/()

y +

2 y +

+ y.
obtain

Equating the two expressions for/(y +

a?),

we

/'(a?)

(a;

az)

(a?

a 3 ) ...

(a;

a) + ....,

as above written,

/"(#)

To- the similar value

The value
of f(x)

of <2_ 2 in terms of x

and the

roots,

may be conveniently written

a2

as follows

73.

multiple Roots.
of the equation f(x)

Theorem.
=
is

multiple root of the

order

a multiple root of the order

0.

m-l

of

the first derived equation f'(x]

This follows immediately from the expression given for/' (x) in the preceding Article for if the factor (x -aO w occurs in /(a?), = a m ; we have i. e. if ai = a 2 =
;
.
. .

a;

ai

x - a m+ i
still

Each term
the
is
first,

in this will

which will have

(x

ai)

have (x - ai) m as a factor, except "1 m ~ as a factor hence z

1

(x

aj"

a factor in

'#.

Determination of Multiple Roots.

COR.
1.

157

Any

root which occurs

m times in the equationf(x] =

m-\

occu rs in degrees of multiplicity diminishing by unity in the first

derived equations.

Since /"(a )
1

is

is

from/(#),

it is

derived from /"(a) in the same manner as f(x] evident by the theorem just proved that (x)
ai)

will contain (x

m ~ 2 as a factor.

The next

derived function,

and so on. m - 1 derived functions all vanish and its first Iff(x) m for a value a of x, then (x - a) is a factor inf(x). This, which is the converse of the preceding corollary, is
/"'(#), will contain (x

a^-

COR.

2.

most readily established

derived functions
substituting in the form

directly as follows
>

Eepresenting the

by /i(#),/2 (#) a + x - a f or x we
9

fm~i(x) (see Art. 6),

find that/(#)

and

may

be expanded

from which the proposition

74.

is

manifest.
It
is

Determination of Multiple Roots.

1

easily

inferred from the preceding Article that if f(x) and/'(#) have a common factor (x - a}" , (x - a) m will be a factor in /(a?) ; for,

by

Cor. 1, the

m-2

next succeeding derived functions vanish

as well as /(a?) and/'(#) when x = a; hence, by Cor. 2, a is a root of /(a?) of multiplicity m. In the same way it appears that if /(a?) and/'(#) have other common factors
(x

13

-1
,

ft}

(x

y) ~\ (x

B) ~\ &C.,
/3,

the equation f(x] = will have r to 7, roots equal to 8, &c.

roots equal to

q roots equal

In order, therefore, to find whether any proposed equation has equal roots, and to determine such roots when they exist,

we must
Let
this

find the greatest

be

0(a?).

common measure of /(a?) and f(x). The determination of the equal roots will
(a?)

depend on the solution of the equation

=0.

158

Properties of the Derived Functions.

EXAMPLES.

1.

Find the multiple roots of the equation

x^

+ x2 is

16z

20

0.

The G.

C.

M.

of f(x) and/'(z)

easily

found to be x

hence (x

2)

is

factor in f(x}.

The

other factor

is

5.

remaining factors,

"Whenever, after determining the multiple factors of/(#), we wish to obtain the it will be found convenient to apply by repeated operations the
division of Art. 8.

method of

Here, for example,

:

we

divide twice

by x

2,

the

calculation being represented as follows

1

1-16 6-20 3-10

2
2

20

10

5
coefficients left, the third factor is

Thus

and 5 being the two

5.

This

operation verifies the previous result, the remainders after each division vanishing as they ought.
2.

Find the multiple

roots,

and the remaining

factor, of the equation

0.

x5 - 10#2 + I5x

-6 =

The G.

C.

M.

of f(x) and f(x)

is

found to be xz - 2x

1.

Hence
obtain

(x

I)

is

factor in f(x).

Dividing three times in succession

by x

1,

we

f(x]
3.

(x

I)

(**

3x

6).

Find the multiple

roots of the equation

12x + 36

0.

x -

The G. C. M. 3. Hence

of f(x)

and/'()
/(*)

is

x*

-x2

6.

The
2
.

factors of this are x

+2

and

(*

2)

(*-3)

4.

Find

all

the factors of the polynomial

f(x)

= z6 -

5* 5

5*4

9z 3

- Uxz - x +
f(x)

8. 1)2

Am.

^( x -l)( x+

X -

2)3.

process of finding the greatest common measure of a polynomial and its first derived function may become

The ordinary

very laborious as the degree of the function increases. It is wrong, therefore, to speak, as is customary in works on the

Theorem.

<

159

of Equations, of the determination in this way of the multiple roots of numerical equations as a simple process, and

Theory

It

one preliminary to further investigations relative to the roots. is chiefly in connexion with Sturm's theorem that the operais

tion

of

any

practical value.

The

further consideration of

multiple roots is deferred to Chap. IX., where this theorem will be discussed. It will be shown also in Chap. X., that the multiple roots of equations of degrees inferior to the sixth can, in

any particular instance, be determined from simple considerations not involving the process of finding the greatest common
measure.

and the succeeding Article will be occupied with theorems which will be found of great importance in the sub75. This

sequent discussion of methods of separating the roots of equations.

Theorem.
a
little less

In passing continuously from a value than a real root a of the equation f(x) =
greater, the polynomials f(x)

a-hofx
to

a value

a + h a

little

signs immediately before the passage through the root,

andf(x) have unlike and like signs

immediately after.

- h in Substituting a /(x) and /(#), and expanding,

we

have

/(a

h)

=f(a) -/(a) h

....,

/(a-A)=
Now,
since /(a)

/(a) -/'(a) h + .....

the signs of these expressions, depending When the sign of h is terms, are unlike. the of the become the same. The changed, signs expressions theorem is therefore proved.
0,

on those of their

first

Corollary.

The theorem remains true when a

is

a multiple

root of any order of the equation f(x)

0.

Let the root be repeated r times.

The

(using suffixes in place of the accents) all

following functions vanish

:

/W,

/i(), /.(a),

-/r-i(a).

160
In the

Properties of the Derived Functions.

series for /(a

h)

and/' (a -

h]

the

first

terms which

do not vanish

are, respectively,

These have plainly unlike signs changed they will have like signs.
established.

but when the sign of h

is
is

Hence the

proposition

76.

Extending the reasoning of the

last Article to

every

consecutive pair of the series

we may state the proposition generally as follows Theorem. When any equation f(x) = has an
:

r-multipk

root a, a value a

little

inferior to a gives to this series of r functions

signs alternately positive

and

negative,

or negative and positive

and a
sign
;

value a

little

superior to it gives to all these functions the

same

and

this sign is,

moreover, the same sign as the sign

offr (a),

the first derived function which does not vanish

when a

is

substituted

for
let

x.

to give a precise idea of the use of this theorem, us suppose that /6 (a) is the first function which does not vanish when a is substituted, and let its sign be negative ;

In order

the conclusion which

may be drawn from the theorem is, that h of x the signs of the series of functions /,/i,/2> for a value a
are

and

for a value a

+ h of x they are

for before the passage through the root the sign of

different
;

must be

from that of /5 the sign of /3 must be different from that of/4, and so on and after the passage the signs must be all the same. It is of course assumed here that h is so small that is included within the interval through no root of f5 (x) =
;

which x

travels.

Examples.

161

EXAMPLES.
1.

Find the multiple

roots of the equation

0.

-4ws. f(x]
2.

(x*

60;

2)

Show

that the hinomial equation

xn
cannot have equal
3.

an

roots.

Show

that the equation

z- nqx + (n will have a pair of equal roots if q n

4.

1)

rn

~l
.

Prove that the equation

x5

5px

+
5

z 5p x

=
if it

has a pair of equal roots when q 2 roots it must have a second pair.
5.

4j?

and that

have one pair of equal

Apply the method of Art.

74, to determine the condition that the cubic

z3

+ 3Hz + G =
common measure must G Z + 4H* = Q.
when
the cubic

should have a pair of equal roots. The last remainder in the process of finding the greatest
vanish.
6.

Ans.
to

Apply the same method

show that both

and

.H"

vanish

has three equal roots.

7.

If a, 0, 7, 5 be the roots of the biquadratic f(x)

0,

prove that

/+/'()8)+/'( 7 )+/'(5)
can be expressed as a product of three factors.

Ans.
8.

(o

-7-

5)

(a

+7of,

5) (a

+ S-

- 7).

If

o, 0, 7, 5,

&c., be the roots of f(x)

0,

and

&,

y',

&c., of /'(#)

prove

and that each is. equal to the absolute term in the equation whose roots are th squares of the differences.
9.

If the equation

X n +piX"- l

+p2X- Z + .... +
is

-\X+n

have a double root a

pi

prove that o
1

a root of the equation

a;"-

2p z z-

Bpzx"-

+ .... + np n =

0.

162
10.

Properties of the Derived Functions.

Show
that the max. and min. values of the cubic
ax*

+ Sbx2 + Sex + d

are the roots of the equation

2

- 2Gp + A =

0,

where

is

the discriminant.

If the curve representing the polynomial /(#) be

(see Art. 10)

moved

parallel to the axis of

through a distance equal to a max. or min. value p, the axis of x will become a tangent to it, *. e. the equation/(#) p = will have equal roots. Hence
the max. and min. values are obtained

by forming the discriminant of f(x)

0.

p,

or

by putting d
11.

p for p in

# 2 + 4H* =

Prove similarly that the max. and min. values of

are the roots of the equation

3

3(

J-

9JET2 ) p 2

+ 3(a/ 2 - 18HJ)

-A=

0,

where
12.

is

the discriminant of the quartic.

of Art. 76 to the function

Apply the theorem

We have

f(x)
/i (x)

= =
=

<c*

7z 3 + 15^2 -

130+4.
13,

4#

2l

+ 30^ 15),

/>

(x)

2 (6#2

- 210 +

/,(*)<= 2 (12* -21),

/*(*)

24.

= 1 and /a (1) is (x) is the first function which does not vanish when x "What the theorem proves is, that for a value a little less than 1 the signs f /> /i> /2> /s are --an d f r a value a little greater than 1 they are all "We are able from this series of signs to trace the functions/, /i, &c., in negative.
Here/s
;

negative.

")

the neighbourhood of the point x

1.

Thus the curve representing f(x)

is

above

the axis before reaching the multiple points = 1, and is below the axis immediately after reaching the point, and the axis must be regarded as cutting the curve in three
coincident points, since (x
to,/i (x) is
3

I)

is

a factor in /(#).

Again, the curve corresponding

.

below the axis both before and after the passage through the point x = 1 The curve representing /2 (x) is above the axis It touches the axis at that point.

before,

and below the axis

after the passage,

and cuts the axis

at the point.

CHAPTER
LIMITS OF
77.

VIII.

THE ROOTS OF EQUATIONS.

In attempting to discover the
it is

Definition of Limits.

real roots of numerical equations,

in the

first

place advan-

tageous to narrow the region within which they must be sought. here take up the inquiry referred to in the observation at

We

the end of Art. 4, and proceed to prove certain propositions relative to the limits of the real roots of equations.

A
them.

superior limit of the positive roots

is

any greater

positive

number than the

tive roots is

greatest of

them

an

inferior limit of the posi-

any smaller

positive

number than the

smallest of

A superior limit of the negative roots is any greater nethem :

gative number than the greatest of them ; an inferior limit of the negative roots is any smaller negative number than the
smallest of

the greatest negative

.

number meaning here

all

that nearest to

oo

When we

have found limits within which

the real roots

of an equation lie, the next step towards the solution of the equation is to discover the intervals in which the separate roots

are situated.

The

principal methods in use for this latter pur-

pose will

form the subject of the next Chapter.

The following Propositions all relate to the superior limits of the positive roots ; to which, as will be subsequently proved, the determination of inferior limits and limits of the negative
roots can be immediately reduced.

78.

Proposition

I.

In any equation

if the first negative

~ term be - p r xn r

and

if the greatest negative

M2

164
coefficient be

Limits of the Roots of Equations.

- pjc,

thenvp* +

is

a superior limit of the positive

roots.

Any

value of x which makes

xn
will,

>pk

n -Jr (x

+ xn ->*~ +
1

+ x+

1)

>p k

r x- 1

a fortiori, make/(#) positive. Now, taking x greater than unity, this inequality
:

is satisfied

by

the following

or
or

xn " - x n

xr

~l

(-.
by the following

which inequality again

is satisfied
1

(x-iy- (0-l)=or>^,
since plainly

xr

~l

>

(x

I)**-

1 .

We have,
or
79.

therefore, finally

# = or >

Proposition II.

ficient be taken positively,

coefficients

If in any equation each negative coefand divided by the sum of all the positive

which precede it, the greatest quotient thus formed increased by unity is a superior limit of the positive roots.

Let the equation be

~ + ---- - a r xn r + ---- + a n =

0,

in which, in order to fix our ideas, we regard the fourth coefficient as negative, and we consider also a negative coefficient in

- ar general, viz.

Let each positive term means of the formula

a mxm = a m (x - 1

in this equation be transformed

by

xm

~l

~ +xm z +

+x+

Propositions.

165

which

is

derived at once from

x"
1

1 1

x-

+ x

the negative terms remaining unchanged. The polynomial/Or) becomes then, the horizontal lines of the

following corresponding to the successive terms of f(x)

+a Q (x- l}x

n~r

+ a*

+ a z (tf-l)^" 3 4
J

+ a,(x-l}xn

-r

+ az ,

-Or*",

We

now regard
&c., being
(a

successive terms in the polynomial

the vertical columns of this expression as the successive coefficients of ;

xn ~\ xn -\

a (x-1),

a,} (x

1),

(a

+ ai + a2 ) (x-l)is sufficient

3,

&c.
positive

Any value

of x greater than unity

to
,

make
,

every term in which no negative coefficient # 3 a r &c., occurs. To make the latter terms positive, we must have

(a

a-i

ffr-i)

(x

1)

> a r &c.
,

Hence
-

1,

&c.

And

to ensure every

term being made

positive,

we must

take

the value of the greatest of the quantities found in this way. Such a value of x, therefore, is a superior limit of the positive
roots.

166

Limits of the Roots of Equations.

The propositions in the 80. Practical Applications. two preceding Articles furnish the most convenient general methods of finding in practice tolerably close limits of the roots. Sometimes one of the propositions will give the closer
limit

It is well, therefore, to apply sometimes the other. both methods, and take the smaller limit. Prop. I. will usually be found the more advantageous when the first negative coef:

ficient is

preceded by several positive

coefficients,

so that r is

large ; and Prop. II. when large positive coefficients occur before In general, Prop. II. will the first large negative coefficient.

give the closer limit.

We

speak of the integer next above the

number given by

either proposition as the limit.

EXAMPLES.
1.

Find a superior limit of the positive #4

Prop.
I.

roots of the equation

5z 3

+ 40z2 - Sx +

23

0.

gives 8

1,

or 9, as limit.

Prop. II. gives

2.

--

!,

or 6.

Hence

is

a superior limit.

Find a superior limit of the positive

x5 +
Prop.
I.
S-r*

roots of the equation

z3
1
;

8* 2

51#

+18 =

0.

gives

y'ol +
1

and 5
1,

is,

therefore, a limit.

Prop. II. gives

-+
-

and 12

is

limit.

In
3.

this case Prop. I. gives the closer limit.

Find a superior

limit of the positive roots of

Of the

fractions

F+V
the third
4.
is

_3

9
1

11

5'

+ 4+5'

6'
I.

the greatest, and Prop. II. gives the limit

3." Prop.

gives 5.

Find a superior limit of the positive roots of

x8

20* 7 + 4

- II* 5 - 120^ + 13# - 25 =

0.

Ans. Both methods give the limit

5.

6.

Find a superior

limit of the positive roots of

9S* 3

- 73* +
I.

0.

Ans. Prop.

gives 20.

Prop. II. gives

3.

Examples.
It
is

167

closer than that given

usually possible to determine by inspection a limit by either of the preceding propositions.

This method consists in arranging the terms of an equation in groups having a positive term first, and then observing what is the lowest integral value of x which will have the effect of rendering each group positive.
the arrangement in
6.

The form

of the equation will suggest

any

particular case.
:

The equation

of Ex. 2 can be arranged as follows

2^.3 _
x=
3,

g)

+x

(3^3

51 )

+ X3 +

18

0.

or any greater number, renders each group positive

hence 3

is

a superior

limit.
7.

The equation
x5

of Ex. 4
3

(x

11)

may be arranged thus + 20*4 (*3 - 6) + 4*6 +

I3x

- 25 =
;

0.
is

3,

or any greater number, renders each group positive

hence 3

a limit.

8.

Find a superior

limit of the roots of the equation

a4

4a; 3

33,^2

23

18

0.

This can be arranged in the form

x2

(x*
2

- 4x +
4x

5)

+ 2Sx(x - A-) +

IS

0.

Now the trinomial x

of x (Art. 12).

+ 5,

having imaginary

roots, is positive for all values

The

a superior limit." introduction in this way of a quadratic whose roots are imaginary, or of one
1 is

Hence x =

with equal
9.

roots, will often

be found useful.

Find a superior

limit of the roots of the equation

5^6

7^4

_ Wx3 _

23^2

_ 90* - 317 =

0.

In examples of

this

kind

it is

convenient to distribute the highest power of x

among

the negative terms.

x* (x

Here the equation may be written

10)

7)

+ *3 (z2 -

+ s 2 (z 3 -

23)

+ x(x* - 90) + z 5 - 317 = 0,

In
this case the general

so that 7

is evidently a superior limit of the roots. give a very high limit.

methods

10.

Find a superior

limit of the roots of the equation

&-x
When
unity,
it

2x*

- 4x -

24

0.

there are several negative terms, and the coefficient of the highest term is convenient to multiply the whole equation by such a number as will

enable us to distribute the highest term among the negative terms. ing by 4, we can write the equation as follows
:

Here, multiply-

x3
and 4
is

(x

4)

+ x 2 (zz The

8)

+x

(x

16)

a:

96

0,

a superior limit.

general methods give 25.

168
81.

Limits of the Roots of Equations.

Proposition III.

Any number

which renders positive

..

the polynomial

f (x}

and all its derivedfunctions fi(x},f<i(x),

.fn (%)

= 0. is a superior limit of the positive roots of the equation f(x] It is much This method of finding limits is due to Newton. more laborious in its application than either of the preceding
methods
limits
;

has the advantage of giving always very close and in the case of an equation all whose roots are real
;

but

it

the limit found in this

way

is,

as will be subsequently proved,

the next integer above the greatest positive root. To prove the proposition, let the roots of the

equation

/ (x) =

be diminished by h

then

x-h = y,

and

If

now h be

such as to

make

all

the coefficients

/m/LW./i (A),..

>/<*)

positive, the equation in y cannot have a positive root ; that is to say, the equation in x has no root greater than h ; hence h is a

superior limit of the positive roots.

EXAMPLE.
f(x]

= x*-

2* 3

Zx*

- Ibx -

3.

In applying Newton's method of finding limits to any example the general mode
of procedure
is
;

as follows

Take the smallest

integral

number which renders

fn-\(x) positive

tuting this
is

and proceeding upwards in order to f\ (x), try the effect of substinumber for x in the other functions of the series. When any function

successively

reached which becomes negative for the integer in question, increase the integer by units, till it makes that function positive and then proceed with
;

the

new

integer as before, increasing

it

again

if

another function in the series

should become negative ; and so on, till an integer is reached which renders all the In the present example the series of functions is functions in the series positive.

/ (x) =

x*

- 2z3 -

3o;2

- 15* -

3,

Inferior Limits,

and Limits of the Negative Roots. 169

We try then the effect of the
Increase
substitution

in/o(jr).

Here x = 1 makes f$ (x) positive. It makes f(x) negative.

;

x=

by

and x

makes fi(x}

positive.
;

Try the effect of x = 2 in f\ (x) it gives a negative result. Increase by 1 and x = 3 makes f\ (x] positive. Proceeding upwards, the substitution x = 3 makes f(x) negative and increasing again by unity, we find that x = 4 makes f(x) posiHence 4 is the superior limit required. tive. It is assumed in this mode of applying Newton's rule, that when any number makes all the derived functions up to a certain stage positive, any higher number will also make them positive so that there is no occasion to try the effect of the on number the functions in the series below that one where our upward higher
;

progress

is

arrested.

This

is

evident from the equation

0(a

h]

<j>(a)

<t>'(a}h

+ $" (a)

A2
-j. .

(taking

<j>(x)

to represent

for derived functions),

and h
It

also positive,

any function in the series, and using the common notation are all positive, which shows that if <(), <?>'( a )> 0"(), must be + <p (a positive. h)
is

may be

observed that one advantage of Newton's method

that often, as in

the present instance, it gives us a knowledge of the two successive integers between which the highest root lies. Thus in the present example, since f(x) is negative for

3,

and positive for x

4.

4,

we know

that the greatest root of the equation lies

between 3 and

Inferior Limits, and Limits of the Negative Roots. To find an inferior limit of the positive roots, the
82.

transformed by the substitution x = -. y Find then a superior limit h of the positive roots of the equation
equation must be
first

in y.
limit
;

The

reciprocal of this, viz. -, will be the required inferior

for since

t,<h,

->-,

!.e.*>-.

To find limits of the negative roots, we have only to transform the equation by the substitution x = -y. This transformation changes the negative into positive roots. Let the superior

and

inferior limits of the positive roots of the equation in

h'.

y be h and

Then - h and -

li

are the limits of the negative

roots of the proposed equation.

170
83.
equation
the

Limits of the Roots of Equations.

Limiting Equations.

If
it

all

the

real roots

of the

'(a?)

=0

could be found,

would

be possible to determine

number of real

roots of the equation f(x)

= 0.

To prove this, let the real roots of/'(#) = be, in ascending order of magnitude, a', /3', y', ... A' and let the following series
;

of values be substituted for x in /(a?)

GO

a, ft,

y',

A',

QO

two of these quantities give results is a root of/(#) =0 between them and by the Cor., Art. 71, there is only one and when they give results with the same sign there is, by the same Cor., no root between them. Thus each change of sign in the results of
successive

When any

with different signs there

the successive substitutions proves the existence of one real root

of the proposed equation. If all the roots of f(x) = are real,
it is

evident,

of Art. 71, that all the roots of /'(a?)

are also real,

by the theorem and that

they

lie

= f(x)

one by one between each adjacent pair of the roots of 0. In the same case, and by the same theorem, it follows

that the roots of /"(a?) = 0, and of all the successive derived functions, are real also; and! the roots of any function lie
severally between each adjacent pair of the roots of the function from which it is immediately derived.

Equations of this kind, which are one degree below the degree of any proposed equation, and whose roots lie severally between each adjacent pair of the roots of the proposed, are called
limiting equations.

evident that in the application of Newton's method of finding limits of the roots, when the roots of /(a?) = are
It
is

proceeding according to the method explained in Art. 81, the function /(a?) is itself the last which will be rendered positive, and therefore the superior limit arrived at is the integer
all real, in

next above the greatest

root.

Examples.

171

EXAMPLES.
Prove that any derived equation /, (x) = cannot have more imaginary roots may have more real roots, than the equation f(x] = from which it is derive". From this it follows that if any of the derived functions be found to havo
1.

but

imaginary

roots, the

same number at

least of

imaginary roots must enter the primi-

tive equation.
2.

Apply the method of Art. 83

to determine the conditions that the equation

x3 - qx
should have
3.
all its roots real.

Determine by the same method the nature of the roots of the equation

xnqx + (n- l)r =

Ans. "When n
q
n
is

0.

even, the equation has

r"- 1
.

two

real roots or none, according as

>

or

<
<

When
q
n

is

odd, the equation has three real roots or one, according as

r"- 1
.

>

or

= has all its roots real hence show, by forming 4. The equation x n (x 1)" the n th derived function, that the following equation has all its roots real and un;

equal,

and situated between

and

_
2n
5.

1.2

2n(2nn th

1)

Show

similarly

by forming the

derived of (x2

l)

n that

the following

equation has

all its roots real

and unequal, and situated between

and

1.2
6.

zero,

If any two of the quantities I, m, n in the following equation be put equal to show that the quadratic to which the equation then reduces is a limiting equa-

tion

and hence prove that the

(

roots of the proposed are all real

X - a}(x-

b)

(x-c}~ I* (x-a)- m*(x -b)- n*(x -c)- limn =

0.

CHAPTER

IX.

SEPARATION OF THE ROOTS OF EQUATIONS.

84.

the methods of the preceding Chapter we are enabled to find limits between which all the real roots of any numerical

BY

equation
to

lie.

Before proceeding to the actual approximation

any

which

it is

particular root, it is necessary to separate the interval in situated from the intervals which contain the remain-

ing roots. The present Chapter will be occupied with certain theorems whose object is to determine the number of real roots

between any two arbitrarily assumed values of the variable. It is plain that if this object can be effected, it will then be possible
to tell not only the total

number of

real roots, but also the limits

within which the roots separately lie. The theorems given for this purpose by Fourier and Budan,

For although different in statement, are identical in principle. of statement is the conFourier's more purposes exposition
of

venient, while with a view to practical application the statement Budan will be found superior. The theorem of Sturm, although
practice, has the

more laborious in
that
it is

advantage over the preceding

number
tities
;

unfailing in its application, giving always the exact of real roots situated between any two proposed quan-

whereas the theorem of Fourier and Budan gives only a

certain limit which the

number

of real roots in the proposed

interval cannot exceed.

85.

Theorem of Fourier and Budan.

of which a
is

Let two numbers

a and

b,

the less, be substituted in the series formed by

f(x] and

its successive

derived functions,

viz.,

Theorem of Fourier and Sudan.

the

173

number of real

roots which

lie

between a and b cannot be greater

the series

number of changes of sign in substituted for #, over the number of changes a stituted for x ; and when the number of real roots
than the excess
/.s-

of the

when
sub-

ichen b

is

in the interval

falls short of that difference,

it

will be by an even

number.

This
It
is

is

the form in which Fourier states the theorem.

to be understood here, as elsewhere, that,

when we
one or
is

speak of two both of them

numbers a and

6,

of

which a
is

is

the

less,
is

may
to

be negative, and what

oo
.

meant

that a

nearer than b to

"We proceed
being supposed

examine the changes which

may occur among

the value of x
b.

the signs of the functions in the above

to increase continuously
:

series,

from a to

The

fol-

lowing
(1).

different cases can arise

The value
0.

of x

may

pass through a single root of the

= equation f(x)
(2). It

may pass through a root occurring r times in f(x] = 0. (3). It may pass through a root of one of the auxiliary functions fm (x) = 0, this root not occurring in either fm .i(x) or /,H+1 (x) = 0. It may pass through a root occurring r times in/m (#) = 0, (4)
.

and not occurring \nfm .^(x) =

venience.

0.
is

In what follows the symbol x

In the
first

omitted after

for con-

from Art. 75, that in passing = one change of sign is through a root of the equation f(x) for / and /i have unlike signs immediately before, and lost
(1).

case

it is

evident,

immediately after, the passage through the root. In the second case, in passing through an r-multiple (2). root of /(#) = 0, it is evident that r changes of sign are lost for,
like signs
;

by Art.
tions

76, immediately before the passage the series of func-

/> /i>

./->

'Jr-1)

Jr

have signs alternately + and -, or - and +, and immediately after the passage have all the same sign as/r
.

174
(3).

Separation of the Roots of Equations.

In the third

case, the root of

fm (x)

must give to/m _!

and/^+i either like signs or unlike signs.

;

Suppose it to give like in the root two then passing through changes of sign are signs of the is the different for before from these fm lost, sign passage

and after the passage it is the same (Art. 76). Suppose it to give unlike signs then no change of sign is lost, for before the passage the signs of /m_i, /m /m+1 must be either
like signs,
; ,

+ +

become + +. On the and we conwhole, therefore, -, can of be variation clude that no sign gained, but two variations may be lost, on the passage through a root of fm (x] = 0. (4) In the fourth case x passes through a value (let us say a) which causes not only/m but also/m+1 ,/m+2 ,/WHr -i to vanish. It is evident from the theorem of Art. 76 that during the passage a number of changes of sign will always be lost. The definite number may be collected by considering the series of functions
,

or

and

after the passage these

Jm-iy

J my fm+lj

J m +r-l>
:

J~m+r'

We easily obtain the following results

(a).

When/

OT _i(a)

and/m+r (a) have

like signs:

If r be even, r changes are lost. If r be odd, r + 1 changes are lost.

(b).

When/w_i(a) and/mr (a)

have unlike signs:

If r be even, r changes are lost. If r be odd, r - 1 changes are lost.

We conclude, therefore, on the whole, that an even number of

changes
is lost

during the passage through an r-multiple root

ofAC*).
It will be observed that (1) is a particular case of (2), and = 1. Since, however, the cases (1) and (3) (3) of (4), i. e. when r are those of ordinary occurrence, it is well to give them a separate classification.

Eeviewing the above proof, we conclude that as x increases from a to b no change of sign can be gained that for each
;

Application of the Theorem.

175

= one change is lost ; and passage through a single root of f(x) that under no circumstances except a passage through a root of = can an odd number of changes be lost. Hence the
;

number

of changes lost during the whole variation of x from = a to b must be either equal to the number of real roots of (x)

in the interval, or

must exceed

it

by an even number.

The

theorem
86.

is

therefore proved.

Application of the Theorem.

The form

in which

the theorem has been stated by Budan is, as has been already observed, more convenient for practical purposes than that just It is as follows Let the roots of an equation f(x) = given.
:

be diminished, first by a

and then by
;

b,

where a and b are any two

bers of which a is the less

then the

n a and b cannot be greater than the changes of sign in the first transformed equation over the number in

number of real roots beexcess of the number of

the second.

This

is

two transformed equations are

'

evidently included in Fourier's statement, for the (see Art. 33)

"

from which, assuming the

proposition
is

results of the last Article, the

above

manifest.

The reason why the theorem in this form is convenient in practice is, that we can apply the expeditious method of diminishing the roots given in Art. 33.
EXAMPLES.
1
.

Find the situations of the roots of the equation *5

3x*

24* 3

95*2

- 46z -

101

0.

We shall examine this function for values of x between the intervals

-10,

-1,

0,

10;

these numhers being assumed on account of the facility of calculation.

Diminution

176
of the roots

Separation of the Eoots of Equations.

by
1,

gives the following series of coefficients of the transformed

equation:
2,

-26,

15,

65,

-78.

In diminishing the roots by 10, it is apparent at the very outset of the calculation that the signs of the coefficients of the transformed equation will be all positive so
;

that there

is

no occasion

to complete the calculation in this case.

In diminishing the roots by 10 and 1, it is convenient to change the alternate signs of the equation, and diminish the roots by -f 10 and + 1 and then in
;

the result change the' alternate signs again. equation when the roots are diminished by
1}

The
1

coefficients of the

transformed

are

_8,

2,

139,

291,

60.

In diminishing by - 10 we observe in the course of the operation, as

before, that

the signs will be all positive in the result, i.e. when the alternate signs are changed they will be alternately positive and negative.

Hence we have

the following scheme

(-10)
'

+ - +
+
-

(-1)
(0)

- + +
-

+
,

+ -

the equation

itself.

(1)

+ + - +

+ -

(10)

+ + + + + +

/3> /4> /5

These signs are the signs taken by/(#) and the several derived functions /i, /2, on the substitution of the proposed numbers but it is to be observed that
;

they are here written, not in the order of Art. 85, but in the reverse order, viz.,
/5,

A, /3, /2, fit / From these we draw

10 and
;

between
of sign

10

the following conclusions one real root lies between

1

All the real roots must lie

10 and

1,

since one change

is lost
;

is lost

one real root lies between

and
1

0, since

one change of sign

no real root lies between

are lost, there
is

and

and between
;

and

10, since three

at least one real root

:

but

we

are left

changes of sign in doubt as to the nature of

real

the other two roots

roots

whether they are imaginary, or whether there are three

10.

between

and

"We might proceed to examine, by further transformations, the interval between and 10 more closely, in order to determine the nature of the two doubtful roots
;

evident that the calculations for this purpose might, if the roots were nearly become very laborious. This is the weak side of the theorem of Fourier and equal, Budan. Both writers have attempted to supply this defect, and have given methods

but

it is

of determining the nature of the roots in doubtful intervals are complicated,

but as these methods

not stop to explain them ; the more especially as the theorem of Sturm effects fully the purposes for which the supplementary methods of Fourier and Budan were invented.

we do

Application of the Theorem

2.

to

Imaginary Roots.

177

Analyse the equation of Ex.

x*

1, p. 100, viz.,
1

+ xz - 1x -

0.

The

roots of this arc all real,

and lie between - 2 and 2

all real,

(see

Ex.

5, p.

100).

When-

ever the roots of an equation are

the exact

the signs of Fourier's functions determine

integers.

numher
:

of real roots

between any two proposed

We obtain the

following result

The

roots lie in the intervals

(-2, -1);
3.

(-1,0);
100, viz.,

(1,2).

Analyse the equation of Ex.

3, p.

Ans.

Two

roots in the interval

(2,
;

1),

and one root in each

;

of the intervals (- 1, 0)
4.

(0, 1)

(1, 2).

Analyse the equation

1998*3

- 14937z +

5000

0.

The equation can have no negative

in succession
till
:

Diminish the roots by 10 several times the signs of the coefficients become all positive. We obtain the
roots.
(0)

following result

i!

178

Separation of the Roots of Equations.

values of x immediately before and immediately after tliat value which causes the coefficient to vanish the whole interval being
;

so small that

it

may
0.

be supposed not to include any root of the

equation f(x)

EXAMPLES.
1.

Analyse the equation

f(x)

=y*- 4z 3 -

3*

23

0.

We shall examine this function "between the intervals 0, 1

equations are
(0)

10.

The transformed

0,

the

first

of these being the proposed equation itself. calculations

0,

Making the
coefficient

/3 (1) =

by the method of the preceding and we have the following scheme

:

Article,

we

find that the

(0)

+ +
--

+
+

(1)

(10)

4-

+ +

by two, the corresponding to a value a little less, and the second to a value a little greater, than that which gives the zero coefficients ; the signs being determined by the It must be remembered that in the above scheme principle established in Art. 76.
first

"We

may now

replace each of the rows containing a zero coefficient

the signs representing the derived functions are written in the reverse order to that The scheme will then stand as follows, using h to repreof the Article referred to.
sent a very small positive quantity
:

(0)

(10)

+ + +

In this scheme the signs corresponding to - h and + h are determined by the condition that the sign of the coefficient which is zero when x = must, when
x
it

h be
t

different

from that next to

it

on the left-hand

side
1

and when x

=+

must be the same. in a similar manner.

The

signs corresponding to 1

- h and

+h

are determined

Rule of
Now

the

Double Sign.
(

179
A,

since a pair of changes is lost in the interval

A),

and since the

equation has no real root hetvreen

h and

A,

we have proved

the existence of a

Two changes of sign are lost between 1 + 7* and 10, so pair of imaginary roots. that this interval cither includes a pair of real roots, or presents an indication of a
pair of imaginary roots.
J.

Which

of these

is

the case remains

still

doubtful.

If several coefficients vanish,

we may he
x*

ahle to establish the existence of

:

several pairs of imaginary roots.

This will appear from the following example

0.

The

signs corresponding to

h and

+h

are,

by the theorem

of Art. 76,

h and + /*, and since 4 changes of sign Hence, since no root exists between are lost in passing from a value very little less than to one very little greater, we are assured of the existence of two pairs of imaginary roots. The other two roots
are in this case plainly real (see Art. 14).

The number of imaginary

this

roots in any binomial equation can be determined in way. 3. Find the character of the roots of the equation

xs

10^3

+ x-

0.

In passing from a small negative to a small positive value of x we obtain the

following series of signs
(-/*)
(0)
:

++

--

+0000+0
+

+ + - + + + +

(+ A)

++

+ -

Since six changes of sign are here lost, there are six imaginary roots. The remaining two roots are, by Art. 14, real one positive, and the other negative.
:

The negative
4.

root lies

betwen - 2 and

c,

and the positive between

and

1.

Analyse completely the equation

There are two imaginary

roots.

Whenever,
it is

as in the present instance, the roots

are comprised within small limits,

In

this

way we

convenient to diminish by successive units. find here a root between and 1, and another between 1 and 2.

we find on diminishing by - 1 that - 1 is itself a root, and writing down the signs corresponding to a value a little greater than - 1, we observe an indication of a second negative root between - 1 and 0.
Proceeding to negative roots,
Analyse the equation

x 5 + y* + x2 - 25*
There are two imaginary roots
;

36

=
-

0.

one real positive root between 2 and 3

and two

real negative roots in the intervals (- 3,

2),

(-

2,

1).

N2

180
88.

Separation of the Roots of Equations.

Corollaries

from the Theorem of Fourier and

of detecting the existence of imaginary

Budaii.

The method

roots explained in the preceding Article is called The Rule of the Double Sign. similar rule, due to De Gua, was in

use before the discovery of Fourier's theorem.

This rule and

Descartes' Rule of Signs are immediate corollaries from the theorem, as we proceed to show.
Cor. 1.

De Gua's Rule for finding Imaginary

:

Roots.

The
cessive

roots ;

rule maybe stated generally as follows When 2m sucterms of an equation are absent, the equation has 2m imaginary and when 2m + 1 successive terms are absent, the equation

has

2m

2, or

2m

imaginary

roots, according as the two terms be-

tween which the deficiency occurs have like or unlike signs. This in as case Art. the number of 85, by examining follows, (4), x lost the of a of from small passage sign during changes negative value

- h to a small positive value

Descartes Rule of Signs.
is
9

h.

Cor. 2.

When

substituted

for

x in the

series

of

functions

/2 (#),/i (#),/(#), the signs are the same as the fn(x),fn-i(%), a Q , a^ a z , the coefficients of 0_i, an , of the proposed signs GO is substituted the signs are all positive. + and when equation
.

Fourier's theorem asserts that the

these limits, viz., the

number

of roots between

number

number
that
is

of variations lost

during the passage from

of positive roots, cannot exceed the to +00,

. .
.

number of changes of sign in the series a , a^ a* an Descartes' rule for positive roots ; and the similar rule for negative roots follows in the usual way by changing the
the
.

This

is

negative into positive roots.

Cor. 3.

Newton's Method of finding Limits.

a number h has been found which renders positive

When

each of the f unctions fn (x),

fn-\(%),

/2(#)> /i(#)> /(#)

since

+ oo also renders each of them positive, it follows from Fourier's theorem that there can be no root between h and + oo that is to say, h is a superior limit of the positive roots; and this is
,

Newton's proposition (Art.

81).

Sturm's Theorem.
89.

181

Sturm's Theorem.
possible

We have already shown (Art. 74)

of a polynomial f (x)

that

it is

by performing the common algebraical operation

of finding the greatest

common measure

derived polynomial to find the equal roots of the = 0. Sturm has employed the same operation for equation f (x) the formation of the auxiliary functions which enter into his
its first

and

method of separating the roots of an equation. Let the process of finding the greatest common measure of The successive ref(.c) and its first derived be performed.
mainders will go on diminishing in degree till we reach '^finally either one which divides that immediately preceding without
remainder, or one which does not contain the variable at all, /. e. which is numerical. The former is, as we have already
seen, the case of equal roots.

The

latter is the case

where no

equal roots exist. It is convenient to divide the discussion of Sturm's theorem into these two cases. shall in the present

We

Article consider the case where no equal roots exist ceed in the next Article to the case of equal roots.

and pro-

The

perclass

formance of the operation itself will of course disclose the to which any particular example is to be referred.

The

auxiliary functions employed

by Sturm

are not the

remainders as they present themselves in the calculation, but the remainders with their signs changed. In finding the greatest

common measure

two expressions it is indifferent whether the signs of the remainders are changed or not in the formation
of
:

of Sturm's auxiliary functions the

change

is essential.

It is

convenient in practice to change the sign of each remainder before making it the next divisor.

Confining our attention for the present, therefore, to the case where no equal roots exist, Sturm's theorem may be stated as
follows
:

Theorem.
in the scries

Let any two real quantities a and b be substituted of n + Ifunctions

'4'uKj

of tic given polynomial f(x)\

its first

derived f^x),

and

182

Separa tion of

the

Roo ts of Equa tions

the successive remainders (with their signs changed) in the process the common measure offinding greatest off(x) andf^x) ; then the
difference between the

is

substituted for x,

expresses exactly the

number of changes of sign in the series when and the number when b is substituted for x number of real roots of the equation f (x) =

between a and

b.

The mode

of formation of Sturm's functions supplies the

following series of equations, in

which

q\ 9
:

q^

qn _i represent

the successive quotients in the operation f(x)

j,/,(*) '-/,(),

These equations involve the theory of the method of finding the greatest common measure for it follows from the first equation that if f(x) and/i(#) have a common factor, this must be
;

a factor

in/2 (#)

like reasoning, that the

and from the second equation it follows, by same factor must occur in/3 (#) and so
;

on,

we come finally and/! (x) have common

till

to the last remainder, which,

factors, will

when/(#)

be a polynomial consisting of these factors. In the present Article, where we suppose the given polynomial and its first derived to have no common
factor, the last

remainder/,^)

is

numerical.

It

is

essential for

the proof of the theorem to observe also, that in the case now under consideration no two consecutive functions in the series

can have a

common

factor

for if they

had we

could,

by reason-

ing similar to the above, show from the equations that this factor must exist also in/(#) and/! (x) ; and such, according to our

In examining, therefore, what hypothesis, is not here the case; can take in the series during the passage of place changes sign of x from a to b, we may exclude the case, of two consecutive
functions vanishing for the same value of the variable
;

and the

Sturm's Theorem.
different cases in

183

which any change of sign can take place are

the following
(1).

When

x passes through a root of the proposed equation

/(*)'= 0:
passes through a value which causes one of the functions /_! to vanish /i,/a , auxiliary x When (3). passes through a value which causes two or
(2).
. . .

When x

more

of the series /,./i,

.fn -\

to vanish together; ,no

two of

the vanishing functions, however, being consecutive.

x passes through a root of f(x) = 0, it follows from Art. 75 that one change of sign is lost, since immediately before
(1).

When

the passage

/()

and/i(#) have unlike signs, and immediately

after the passage they have like signs. x to take a value a which satisfies the equation (2). Suppose

fr(%)

0.

From

the equation

/r-l(tf)=?r/r(#)-/mM

we have

/r-i(a) =-/r+i(o),

which proves that this value of x gives to/ r-i(#) and/ r+ i(a?) the same numerical value with different signs. In passing from a
value a
little less

than a to one a
it

little

greater,

the interval so small that

contains no root

we can suppose of/,._i(o?) or/r+1 (#)

;

hence, throughout the interval under consideration, these two functions retain their signs. If the sign of fr (x] does not change

happen in the exceptional case when the root a is rean even number of times) there is no alteration in the peated series of signs. In general the sign of fr (x) changes, but no
(as will

variation of sign is either lost or gained thereby in the group of three ; because, on account of the difference of signs of the two

extremes fr-\(x)

and/r+ i(#),

there will exist both before, and after

the passage one variation and one permanency of sign, whatever be the sign of the middle function. If, for example, before the
i. e.

after the passage they are + + -, passage the signs were + a variation and a permanency are changed into a perma; ;

nency and a variation but no variation of sign on the whole.

is lost

or gained

184

Separation of the Roots of Equations.

(3). Since the reasoning in the previous cases is founded on the relations of the function to those adjacent to it only and
;

since those relations

remain unaltered in the present

case, be-

that if/ (a?)

is lost,

cause no two adjacent functions vanish together, we conclude is one of the vanishing functions, one change of sign

and if not, no change is either lost or gained. "We have proved, therefore, that when x passes through a root of /(#) = one change of sign is lost, and under no other
is

circumstances
the

a change of sign either lost or gained.

Hence

number

of changes of sign lost during the variation of x

from a

to b is equal to the

number

of roots of the equation be-

tween a and b*
Before proceeding to the case of equal roots,

we add

a few

simple examples to illustrate the application of Sturm's theorem. It is convenient in practice to substitute first - oo , 0,

oo

in Sturm's functions, so as to obtain the whole

number

of

negative and of positive roots. To separate the negative roots, - 1, - 2, - 3, the &c., are to be substituted in succession
integers
till

we reach the same

;

tion of - oo

and

series of signs as results from the substituto separate the positive roots we substitute

1, 2, 3, &c., till

the signs furnished

by +

oo are reached.

EXAMPLES.
1.

Find the number and situation of the real roots of the equation
/(*)

We find

/!

(x)

Zx*

2,

= *3 - 2x - 5 = 0. /2 (z) = 4^+15,
0,

fz (x] = -

643.

Corresponding to the values

oo

+
-

oo

of x,

we have
,

(-00)
(0)

I
;

+ + + +
-

(+

oo

.+
and

-.;

Hence there is only one

*

real root,

it is positive.

The student

changes of sign can be

often finds a difficulty in perceiving in what way a number of lost in Sturm's series, since the only loss of sign takes place

between the

It may tend to remove this diffifirst two functions, f(x) and f\ (x). = to a second culty to observe, that as x increases from one root a of f(x) although no alteration takes place in the number of changes of sign, the distribution
,

of the signs

among /i (x} and

the following functions alters in such a

way

that the

signs of f(x)

through

a,

and f\ (x), which were the same immediately after the passage of x become again different immediately before the passage through j8.

3?

186
"We
obtain,

Separation of the Roots of Equations.

removing the factor 2 from the derived,
/i
(x)

= 2* 3 -

3* 2

3s +

5,

/z (s)=9**-27*+ll, /(*) = 8s- 3,

/4
(*)

=-1433.
it is

[N.B.
equations

In forming Sturm's functions

(1),

allowable, as

is

evident from the

Art. 89, to introduce or suppress numerical factors just as in the

;

process of finding the G.

c. M. taking care, however, that these are positive, so that the signs of the remainders are not thereby altered.] "We have the following series of signs
:

(-00)
(0)

+.+
+ +

+
-

_,

-,

(+00)

+ + +
and one negative and two imaginary and

Hence
roots.

there are

two

real roots, one positive,

To

find the position of the real roots, it is sufficient to substitute positive

and negative
2 and
3,

integers successively in f(x) alone, since there is only one positive

:

one negative root

we

easily find in this

way

that the negative root lies between

1.

and the positive root between

and

90.

Sturm's Theorem.

Equal Roots.

Let the opera-

tion for finding the greatest common measure of/(#) and/'(#) be performed, the signs of the successive remainders being

changed as before.

The

last of

Sturm's functions will not

now

be numerical, for since f(x) and/('#) are here supposed to contain a common measure involving #, this will now be the last
function arrived at

by the

process.

Let the

series of functions

be:

During the passage

of x through

any value except a multiple root

still true with no value except such a root

of f(x) = 0, the conclusions of the last Article are

respect to the present series, since

When x = passes through a multiple root of/(#) 0, there is, by the Cor., Art. 75, one change of sign lost between/ and / and we procan cause any consecutive pair of the
series to vanish.
x ;

ceed to prove that no change of sign is lost or gained in the rest of the series, viz./,/., .fr Suppose there exists an m-mul.
. .

tiple root a of/(a?). It

is

evident from, the equations

(1)

of Art. 89,

Sturm's Theorem.
~
. .
.

187

l that (x - a) m is a factor in each of the functions/!, /, ./,.. Let the remaining factors in these functions be, respectively,

By dividing each of the equations (1) by a series of equations which establish by the (x a)'"' get the last Article that, owing to a passage through a, reasoning of no change of signs is lost or gained in the series 0i, 2 0,..
0!,
0e,
.
.

r.

we

any change gained in the series for the effect of the factor /;, /:, /,. ; (x a)'"- in the passage of x from a value a - h to a value a + h is either to change the
Neither, therefore,
.
. .

is

lost or

signs of all (when of the functions 0i, these

m2,

-is

odd) or of none (when m 1 is even) and changing the signs of all $/

;

functions cannot increase or diminish the

number

of

variations.

have therefore proved that when x passes through a = one change of sign is lost between / and multiple root of f(x]

We

/n

and none

either lost or gained in

any other part of the

series.

It remains true, of course, that

when x

root of/ (x) =

a change of sign is lost the state theorem as follows for the case of equal roots

passes through a single as before. "We may thus

:

The

difference between the

number of changes ofsign when a and

are substituted in the series

the

last

equal

to the

of these being the greatest common measure off and /, is number of real roots between a and b, each multiple

root counting only once.

EXAMPLES.
1
.

Find the nature of the

a4

roots of the equation

bx*

9z 2 - 1x

0.

"We easily obtain

/i(*)

4* 3

15z 2

+ 18* -7,
;

/(*)

= **- 2* + 1
;

fz (x) divides f\ (x) without remainder

hence in

this case

Sturm's series stops at

/>(#), thus establishing the existence of equal roots.

188
To
find the

Separation of the Roots of Equations.

number
of real roots of the equation,

we

substitute

oo

and

<

for x in the series of functions/, /i, f%.

The

result is

(-co)
(+
oo
)

+
+

- +, +
+
.

Hence the equation has only two

as
is

real distinct roots

(x),

but one of these

2

is

a triple root,

evident from the form of fz

which

is

equal to (x

I)

2.

Find the nature of the

x*

roots of the equation

6* 3

13*2

12*

0.

Here
/i
(x)

= 4*3 2

18*2

26*

12,

/2 (*)=* -3* +
fz(x)
is

2;

the last Sturmian function

so the equation has equal roots.

(-co)
(+
co
)

+,
.

+ + +
In

There are only two real

the roots
3.
1,

distinct roots.

fact, since

/2 (x) = (x -

l}(x

- 2),

each of

is

a double root.
roots of the equation

Find the nature of the

x5

2* 4

+ x* - xz - 2* -f

0.

Here
/i

= 5* 4 + 8* 3
2*3

3*2 - 1x

2,

/2 =

7*

+
-

12*
5,

7,

/3 = - * 2

60;

Since /g =

0,

1 is

common measure

of /and /i,

and/(#) has a double root

1.

"We have

also

(_,)
(+00).

_ +
+

__+,

+ +

there are two real distinct roots. The equation has, therefore, beside the double root, one other real root, and two imaginary roots.

Hence

4.

Find the nature of the

*
G

roots of the equation

5

7*

+
6*
5

15* 4

- 40*2 + 48* -16 =

4

0.

Here
/i (x)
fz(x)

35*

60*

80*

48,
48,

= 13* 4 - 84* 3 + 192*2 - 176* +

/3 (x)
Am.

*3

6*2

12*

(*

2)

There are three real distinct

roots,

one of them being quadruple.

Application of Sturm's Theorem.

91.

189

Application of Sturm's Theorem.

In the case of

equations of high degrees the calculation of Sturm's auxiliary It is important, therefunctions becomes often very laborious.
fore, to

what

to

(1).

pay attention to certain observations which tend somedimmish this labour. In calculating the final remainder when it is numerical,
is all

since its sign

we

are concerned with, the labour of the last

operation of division can be avoided by the consideration that the value of x which causes n _i to vanish must give opposite

It is in general possible to tell without any signs tofn-z and /. calculation what would be the sign of the result if the root of
/n-i(*)

were substituted
o
~,

mfn^(x).

Thus

in

Ex.
0,

3, Art. 89, if

the value -

which

is

the root of

o for x in 9ar - 27 x + 11, the result

3 (a?)

be substituted

is evidently positive; hence the sign of fn (x) is -, and there is no occasion to calculate the value - 1433 given for/n (a?) in the example in question.

(2).

When

it is

possible in

any way to recognize that

all

the

roots of

any one of Sturm's functions are imaginary,

we need
;

not proceed to the calculation of any. function beyond that one

for since such a function retains constantly the

same sign

for all

values of the variable (Cor. Art. 12), no alteration in the number of changes of sign presented by it and the following functions

can ever take place, so that the difference in the number of

changes when two quantities a and b are substituted is independent of whatever variations of sign may exist in that part of
the series which consists of the function in question and those following it. With a view to the application of this observation

always well, when we arrive at the quadratic function + bx + c, suppose), to examine, in case the term containing (ax x~ and the absolute term have the same sign (otherwise the roots could not be imaginary), whether the condition 4ac > b~ is fulit

is
2

filled

if so,

we know

that the roots are imaginary, and the cal-

culation need not proceed farther. Similar observations apply to the case where one of the

functions
its

is a perfect square, since such a function cannot for real values of x. sign

change

190

Separation of the Roots of Equations.

EXAMPLES.

Analyse the equation

x*

3* 3

We find
ft (x)

7*2

10*

0.

= -29*2-78* +14,
1086*

MX) =-

481,

Here we see immediately that the value of x given by the equation /3 (x) = 0, which differs little from hence fi(x) is negative. |, makes fz(x] positive There are two real, and two imaginary roots. The real roots lie in the intervals
;

{-2,-!},
2.

{-1,0}.
#4

Analyse the equation

4^3

3a;

We find
fz(x)

23

0.

12*

9*

89,

/3

(a;)

= -491*;+

1371,

/*(*)=
Here

-^>
>
2-74

/3 (*) =

gives

a;

>

-,

and x

- makes

fz(x) positive ; hence the root of /a (a?) makes it positive also. There are two real and two imaginary roots.

The
3.

real roots lie in the intervals

{2,3},

{3,4}.
19

Analyse the equation

2s4

13* 2

10*

=
5,

0.

Here

fi(x)

4* 3

- 13* +
loa;

fz(x)

= 13*2-

+ 38.
we proceed no

Since

4 x 13 x 38

>

15 2 ,

the roots of/2() are imaginary, and

farther with the calculation of Sturm's remainders.

Substituting

oo

0,

+00,

we

obtain

(-QO)
(0)

+
+

+, +,

(+00)
There are two
4.

+.

real roots,

one positive, the other negative.

Analyse the equation

/(*)

x5 + 2* 4

+
+

*3

4z 2

3* - 5
3,

0.

Here

MX) = 5s4 + /3 (*) = 6*3 +

/3 (*) = -

8*3

3*

- 8* -

66*2

116*2

+ 44^+119, -57* -223.

Examples.
Since 4 x 116 x 223
co
,

191

>

57-,
,

we may

stop the calculation here.

We

find,

on

ituting

0,

-f oo

(-00)
(0)

-,
-.

(+00)
There are four imaginary
5.

roots,

and one

real positive root.

real roots of the equation
0.

Find the number and situation of the

j*

2j,-3

7.^2

+ 10* +10 =
all real,
1,

Ans. The roots are

and are situated in the intervals

(-3, -2}, {6.

0}, and

two between

{2, 3}.

Analyse the equation

It will

3* 2

- 2* -

0.

be found that the calculation

may

cease with the quadratic remainder.

:

Ans. There
7
.

is

only one real root

in the interval {1, 2}.

Analyse the equation

We find

/2 (x) =

854*

27.51,

In some examples, of which the present is an instance, it is not easy to tell immediately what sign the root of the penultimate function gives to the preceding We have here calculated fz(x), and it turns out to be a much smaller function.

number than might have been expected from the magnitude of the

coefficients

in/2 (a;)

In fact when the root of/2 (#) is substituted in f\ (x) the positive part is nearly equal This is always an indication that two roots of the proposed to the negative part. There are in the present instance two positive roots beequation are nearly equal.

tween 3 and
3*2

4.
;

Subdividing the intervals,

we

find the

two

roots

still

to lie

between

and

3'3

so that they are very close together.

We

see here another illustra-

tion of the continuity

which exists between real and imaginary roots. If /3 (x) turned out to be zero, the roots would be actually equal. If it turned 0'it to be a small negative number, the two nearly equal roots would be imaginary.
8.

Analyse the equation

The

quadratic function

is

found to have imaginary

roots.

Ans. One real root between (0, 1}; four imaginary roots.

192
9.

Separation of the Roots of Equations.

Analyse the equation
y*

Q X5

30

We find
and
as this has plainly all imaginary roots, the calculation

may

stop here.

Ans.
10.

Two

real roots

in the intervals

{2,
5

1),

{6, 7).

Analyse the equation

ISx

0.

We find
/2 (x)
and the calculation may
stop.

= 5z 4

Ans.
11.

Two real

roots

in the intervals

{-1,0},

{5,6}.

Examine how the

roots of the equation

2z 3

15* 2

Ux -

190

=
oo
,

are situated in the several intervals hetween the

numhers
14,

7, 6,

oo

Here

/i

(a;)

z?

+ 5x 40,

/2 (*)

= 27# +

AIM
The

substitution of the ahove quantities gives

(-GO)
(-7)
(6)

+
+

+,
+,

+ +

+ +

+,

(+ oo

+
=

+
may

Whenever,
vanish (here

as in this example,

any quantity makes one

0),

of the auxiliary functions

he disregarded in counting the number of changes of sign in the corresponding row for, since the signs on each
7 satisfies f\(x)

the zero

side of it are different,

no

alteration in the

number

of changes of sign in the

row

could take place, whatever sign be supposed attached to the vanishing quantity. The roots are all real. There is one root between oo and - 7 and two be;

tween
12.

7 and 6.

Analyse the equation

3* 4

6s2

- 8x -

0.

We find
/i(*)

= 3*-3*-2,
8
'

/2 (*)=(*+

I)

As /a (x)

is

a perfect square the calculation

may

cease.

Ans.

Two

real roots

in the intervals

{-1,

0},

{1, 2}.

Conditions for Reality of Roots of an Equation.

92.

193

f(x)jf'(x)

Conditions for the Reality of the Roots of an Equation. The number of Sturm's functions, including and the n - 1 remainders, will in be n + 1.
general
certain cases,

In

function,

some

to the absence of terms in the proposed of the remainders will be wanting. This can

owing

occur only
it is

when the proposed equation has imaginary roots

for

plain that, in order to insure a loss of n changes of sign in the series of functions during the passage of x from - oo to + oo

(namely, in order that the equation should have all its roots And, moreover, they real), all the functions must be present. and alternating must all take the same sign when x = + oo
;

Since the leading term of an equation is signs a with taken positive sign, we may state the condition always for the reality of all the roots of any equation (supposed not to
oo
.

when x = -

have equal

In order that all the roots of an roots) as follows th should be real, the leading coefficients of n the of degree equation all Sturm's remainders, in number n - 1, must be positive.
:

EXAMPLES.
1.

Find the condition that the

roots of the equation

ax*-

2bz +

should be real and unequal.

2.

Am. b*-ac>
z3

0.

Find the conditions that the roots of the cubic

+ 3Hz + G =
from

should be

all real

and unequal.

"When
which

this cubic has its roots all real, it is evident that the 'general cubic

derived (Art. 36) has also its roots all real ; so that, in investigating the conditions for the reality of the roots of a cubic in general, it is sufficient to discuss
it is

the form here written.

We find
f (z)=z* + H, fz (z)=-2Hz-G,
l

[In calculating these, before dividing f\

2 positive factor 2.H" .]

(z)

by/2 (z),
and

multiply the former by the

Hence the required

conditions are,
as

.H" negative

These can be expressed implies the former (cf. Art.

one condition,

viz.,

G z + 4H 3 negative. G z + 4 JET 3 negative, since

this

43).

194
3.

Separation of the Roots of Equations.

Calculate Sturm's remainders for the biquadratic
z4

6J2z2

+ 4Gz +

a*I -

ZH* =

0.

We

find

/2

(*)

= - 3#s2 -

3z

- 3B" 2 ),. (a*I

These are obtained without much

Before dividing fi by

difficulty

by

aid of the identity of Art. 37.

;

fa multiply by

the positive factor 3JET 2

.

and when the re-

mainder

is

2 found, remove the positive factor a

by the

positive factor

(2HI

3a/)

2
;

and when

Before dividing fz by/3, multiply the remainder is found, remove the

positive factor a?H*.

Conditions for the Reality of the Roots of a Riquadratic. In order to arrive at criteria of the nature of
93.

the roots of the general algebraic equation of the fourth degree by Sturm's method, it is sufficient to consider the equation of

Ex. 3 of the preceding

Article.

By

aid of the forms of the

leading coefficients of Sturm's remainders there calculated, we can write down the conditions that all the roots of a biquadratic
should be real and unequal in the form

negative,

2HI- 3aJ negative,

P - 27 J

positive.
is

It will be observed that the second of these conditions

different in

form from the corresponding condition of Art. 68. the To show equivalence of the two forms it is necessary to prove that when If is negative and A positive, the further conthe condition a~I - 12IF dition 2HI - 3aJ
negative implies
negative, and conversely. written in the form it

From

the identity
2

H(a?I- 12H
when
is

readily appears that

is

H and

Art. 37, 2 a*(2HI-3aJ) - 3 ,

of

2HI - 3aJ

are negative

a* I - 12H*

necessarily negative.

And

to prove the converse

we

observe that

when aJ

positive

2HI

- 3aJ

is

negative,
;

/ is positive on account of the condition A positive and when aJ is negative 2HI - 3aJ is still negative, since the - 3aJ as may be negative part 2HI exceeds the positive part readily shown by the aid of the inequalities 12J? > a l and
since
9

P > 27 J\
The student will have no difficulty in verifying, by means of Sturm's functions, the remaining conclusions arrived at in the different cases of Art. 68.

Examples.

195

EXAMPLES.
1
.

Apply Budan's method

t

to separate the roots of the equation

x _ 16*3

69^2

70^

-42 =
1,

0.

Ans. Roots in intervals {2.

0}, {2, 3}, {4, 5}, (9, 10}.

Apply Sturm's theorem to the analysis of the equation

at

3?

Itf

6x

0.

In analysing a biquadratic of this nature which has plainly two real roots, when a Sturmian remainder is reached whose leading coefficient is negative, the calculation

may

cease, since the other pair of roots

must then be imaginary, and the

of the real roots can be readily found

by
;

Ans.
3.

Two

roots imaginary

positions substitution in the given equation. real roots in intervals {- 1, 0}, (2, 3}.

Analyse in a similar

manner the equation

x*-ox? + lOz2 - Qx - 21 = 0.
Ans.
4.

Two

roots imaginary

real roots in intervals

{-1, 0},

{3, 4}.

Apply Sturm's theorem

a;4

to the analysis of the equation

3x 3

x-

- 3x+ll =

0.

Ans. Roots
5.

all

imaginary.

Find by] Sturm's method the number and position of the

real roots of the

equation

Ans. Roots

all real;

{1,
6. If, in

one in the interval {-4, - 3[; two in the interval 0}; and positive roots in the intervals {0, 1}, {3, 4}.

ficients of

the following, the sequences of [signs are those of the leading coefSturm's remainders for a biquadratic, prove

+ + +

four real roots

>

two

real roots

no

real root

cannot occur.

7. If .the

signs of the leading coefficients of the first

two of Sturm's remainders

is

for a quintic be

+ prove
,

that the

number

of real roots

determined.
real root only.

Ans. One
8.

If

H and /are both positive,

;

imaginary

the roots of the biquadratic are and that under the same conditions the quintic written with binomial

prove that

all

coefficients

has [only one real root.

Mr. M. Roberts, Dublin Exam. Papers, 1862.

196

Separation of the Roots of Equations.

if c

t$. Prove that,

has any value except unity, the equation

has a pair of imaginary roots.

10.

Prove that the roots of the equation

are all real, and solve

-

it

when two

of the quantities a,

b, c

become equal.

11.

Prove that when the biquadratic

f(x)

ax*

has a triple factor,

it

may

be expressed in the form

a*f(x}
12. Verify
filled

{ax

+ VTff} 3 {ax +

V^Tff}

by means

of Sturm's remainders the conditions

is

which must be ful-

when

the biquadratic of the previous example

a perfect square, and prove

in that case

13. If an equation of any degree, arranged according to powers of x, have three consecutive terms in geometric progression, prove that its roots cannot be all real. ~ These three terms must be of the form kx r + kax r l kaz xr -*. Let the equation
\-

be multiplied by x
absent, and

a.

The

resulting- equation will

have two consecutive terms

roots
;

must therefore have

at least

two imaginary

but

all

the roots of

this equation except a are roots of the given equation.

14. If an equation have four consecutive coefficients in arithmetic progression, prove that its roots cannot be all real. This can be reduced to the preceding example. Writing down four terms of the
1, it readily appears that the resulting equation proper form, and multiplying by x has three consecutive terms in geometric progression.

CHAPTER

X.

SOLUTION OF NUMERICAL EQUATIONS.

94.

Algebraical and Numerical Equations.

There

is

an

essential distinction

numerical equations. mula of a purely symbolical character, which, being the general expression for a root, must represent all the roots indifferently.
It

between the solutions of algebraical and In the former the result is a general for-

must be such
it

that,

when

for the functions of the coefficients

the corresponding symmetric functions of the roots are substituted, the operations represented by the radical

involved in

and when the square and signs ^/ y/ become practicable cube roots of these symmetric functions are extracted, the whole expression in terms of the roots will reduce down to one root
?
;
:

the different roots resulting from the different combinations

y/ of square roots, and y/, For a simple illustration of what

case of the quadratic in Art. 55. similar illustrations for the cubic

<o

y/,

w 2 y/

of

cube

roots.

is

here stated

we

refer to the

observed, also,

In Articles 59 and 66 we have and biquadratic. It is to be that the formula which represents the root of an

algebraic equation holds good even

when

the coefficients are

imaginary quantities. In the case of numerical equations the roots are determined and, before separately by the methods we are about to explain
;

attempting the approximation to any individual root, it is in general necessary that it should be situated in a known interval

which contains no other real

root.

The

real roots of numerical equations

;

may

be either com-

mensurable or incommensurable

the former class including and fractions, integers, terminating or repeating decimals which

198

Solution of Numerical Equations.

the latter consisting of interminable roots of the former class can be found exactly,
;

are reducible to fractions

decimals.

The

and those
accuracy,

of the latter approximated to with

any degree of

We
1

shall

by the methods we are about to explain. commence by establishing a theorem which reduces

the determination of the former class of roots to that of integral

roots alone.

95.

Theorem.
is

An
and

equation in which the coefficient of the

the coefficients

fast term

of the other terms whole numbers, cannot have a commensurable root ivhich is not a whole number.
unity,

For,

if possible, let -,

a fraction in

its

lowest terms, be a

root of the equation

xn + p l xn

-1

we have then

from which, multiplying by

n
~*,

we

obtain

Now

is not divisible by b, and each term on the right-hand the side of have, therefore, a fracequation is an integer. tion in its lowest terms equal to an integer, which is impossible.

an

We

Hence - cannot be a

root of the equation.

The

real roots of

the equation, therefore, are either integers or incommensurable

quantities.

numbers, fractional or not, can be reduced to the form in which the coefficient of .the first term is unity and those of the other terms whole
so that in this way, by the aid of a simple determination of the commensurable roots the transformation, in general can be reduced to that of integral roots.

Every equation whose

coefficients are finite

numbers

(Art. 31)

Newton's Method of Divisors.

199

proceed to explain Newton's process, called the Method of Divisors, of obtaining the integral roots of an equation whose
coefficients are all integers.

We
96.

lewton's Method of Divisors.

Suppose h to be an

integral root of the equation

of

'1

+ .... + a n .\ x + a n =
is

0.

(1)

Let the quotient, when the polynomial

bQ

divided by x -

h,

be

xn

~l

bi sf

'2
-f
. . .

b n_2 x

bn-i 9

in which b

b^ &c., are plainly all integers. Proceeding as in Art. 8, we obtain the following equations
,

tfn-2

n _z

n _z ,

#_! =

n_i

n ^,

an = -

^i.

The last of these equations proves that a n is divisible by /*, The second last, which is the same as the quotient being - b^i.
a n -\

an
-r

= - hbn_^
and the
se-

proves that the

sum

of the quotient thus obtained

cond
_2

last coefficient is
;

again divisible

by

h,

the quotient being

and

so on.

Continuing the process, the last quotient obtained in this way will be & 0> which is equal to a
.

If

we perform

sors of a n

the process here indicated with all the diviwhich lie within the limits of the roots, those which

satisfy the above conditions, giving integral quotients at each

,

and a

final quotient

-a equal to

are roots of the proposed

Those which at any stage of the process give a fracequation. tional quotient are to be rejected.

When
all

the coefficient a =
()

1,

we know by

the theorem of the

last Article that the integral roots

determined in this

way

are

the commensurable roots of the proposed equation.

If a Q be

200
not=
1,

Solution of Numerical Equations.

the process will still give the integral roots of the equastands ; but to be sure of determining in this way all

tion as

it

the commensurable roots, the equation must be. first transformed to one which shall have the coefficient of the highest term equal
to unity.

97. Application of the Method of Divisors. With a view to the most convenient mode of applying the Method of Divisors, we write the series of operations as follows, in a manner analogous to Art. 8
:

-b

-bi

-b

The
dividing

first
fl

n |by h.

figure in the second line (- & M _i) is obtained by This is to be added to a,^ to obtain the first

figure in the third line (-

hb^).

This

is

to be divided
;

by h

to

obtain the second figure in the second line (- b n_2) this to be added to a n ~z ; and so on. If h be a root, the last figure in the

second line thus obtained will be - a Q

any integer a root, the next operation with any divisor may be performed, not on the original coefficients a n) a^i, but on those of the ,
is
.
.
.

When we succeed in proving in this manner that

.

second line with their signs changed, for these are the coefficients of the quotient when the original polynomial is divided by x - h.

When any
stopped.

divisor gives at

any stage a

fractional result
it is

it is

to

be rejected at once, and the operation so far as

concerned

The numbers
divisors of a n ,

1, which are always of course integral need not be included in the number of trial divisors.

and -

is more convenient before applying the Method of Divisors to determine by direct substitution whether either of these numbers is a root.

It

Examples.

201

EXAMPLES.
1
.

Find the integral roots of the equation

a*

2x?

13z2

38.r

24

0.

By grouping
roots lie

between

the terms (see Art. 79) we observe without difficulty that 5 and + 6. The following divisors are possible roots
:

all

the

-4,
"We commence with 4
:

-3,

-2,

2,

3,

4.

-.24

38

-13
8

-2

-6
32

The

operation stops here, for since

3

5 is not divisible
:

by

4,

4 cannot be a root.

"We proceed then with the number

24

38

13

10

-1
.

30

-3

0;

hence 3

is

a root

and in proceeding with the next

coefficients of the

integer, 2,

we make
:

use, as

above explained, of the

second line with signs changed

1

8-10
4
-

-3 -2
2
:

-1
0;

hence 2 also

is

a root

and we proceed with

-4

2 is not a root, for it does not divide 5.

4.

is

plainly not a root, for

it

does not divide

[We might
S

at once

have struck out

of the reduced polynomial. the number of dh

'

3 as not being a divisor of the absolute This remark will often be of use in diminishing

202
"We proceed

Solution of Numerical Equations.

now
to the last divisor,

4
3
1

-4

Thus 4. is a root. The equation has,

therefore, the integral roots 3, 2,

and the

last stage

of

the operation shows that when the original polynomial is divided by the binomials x 1 ; so that 1 is also a root. Hence the original 3, x 2, x + 4, the result is x

polynomial

is

equivalent to

(*-!)(*2.

2)

(*-3)(*+

4).

Find the integral roots of

Zlx

30

0.

The
2,
3,

roots lie
6.

between

2 and 8

hence

we have

only to

test the divisors

5,

We find immediately that 6 is not a root.

For 5 we have

-30
-

31
6

35-23
5

8-3
;

25

40

15

hen ce 5

is

a root.

For

6-5-8
2

we have

-3
0;

-3
hence 3
is

-9

a root

and we

easily find that 2 is not a root.

(x

The

quotient,

when the original polynomial is divided by

3*
2

5) (x

3), is,

from

the last operation,

+x-

of this 1 is not a root,

and -

1 is

a root.
2

Hence

all

the integral roots of the pro-

posed equation are

1, 3, 5. is -.

The

other root of the equation

It is a

commensurable root

but, not being

integral, is not given in the above operation.

3.

Find

all

the roots of the equation

#4
Limits of the roots are

x3

2x-

+ 4# -

24

0.

4, 3.

Am.

Roots

3, 2,

\/~.

Method of Limiting
4.

the

Number of Divisors.

203

Find

all

the roots of the equation

x*

2*3

- 19*2 + 68* 6.

60

0.

The

roots

He between

- 6 and

We
-

find that 2, 3,

5 are roots, and that the factor left after the final division a double root.

hence 2

is

The polynomial

is

therefore equivalent to

In Art. 99 the case of multiple roots will he further considered.

98.

Method of Limiting the X umber of Divisors.

of course to

It is possible

determine by direct substitution

whether any of the divisors of a n are roots of the proposed equabut Newton's method has the advantage, as the above tion
;

examples show, that some of the divisors are rejected after verylittle labour. It has a further advantage which will now be
explained. of the roots

When the number

is
it is

of divisors of a n within the limits

important to be able, before proceeding large, with the application of the method in detail, to diminish the number of these divisors which need be tested. This can be

done as follows
If h
is

an integral root of/(#) =0,/(^)

is divisible

and the

coefficients of the quotient are integers, as

by x h, was above

If therefore we assign to x any integral value, the explained. quotient of the corresponding value of/(#) by the corresponding value of x h must be an integer. "We take, for convenience,

we
1

the simplest integers 1 and - 1 and, before testing any divisor /*, subject it to the condition that /(I) must be divisible by
;

1) changing the sign, by h be divisible by h (or, changing the 1

(or,

- h

and that/(- 1) must sign, by 1 + h}.

found convenient to
if

In applying
calculate /(I)

this observation
1) in the

it

will be

and/(-

first

instance:

either of these

vanishes, the corresponding integer is a root, and we proceed with the operation on the reduced polynomial whose coefficients

have been ascertained in the process of finding the result of substituting the integer in question.

204

Solution of Numerical Equations.

EXAMPLES.
1.

x 5 - 23# 4
roots lie

160*3
24.
:

281* 2

- 257* -

440

0.

The

between

and

We have the following divisors

2,
4,

5,

8,

10,

11,

20,

22.

We
We

easily find

/(I)

=-

840,

and /(-

1)

=-

648.

therefore exclude all the above divisors, which,

;

when diminished by

1,

do

and which, when increased by 1, do not divide 648. The first condition excludes 10 and 20, and the second 4 and 22. Applying the Method of
not divide 840
Divisors to the remaining integers 2, 5, 8, and that the resulting quotient is #2 + x +
valent to
11,
1.

we

find that 5, 8,

and 11 are
is

roots,

Hence the given polynomial

equi-

2.

x5
roots lie

29* 4

31*3

31*2

32*

60

0.

The

between

2,

and
3,

32.

Divisors

so 1
is

2,

4,

5,

6,

10,

12,

15, 20,

30.

/(I)

=
1)

a root.

/(-

124

We easily find
99.

and the above condition excludes all the divisors except - 2, 3, 30. that - 2 and 30 are roots, and that the final quotient is xz + 1
is

The given polynomial

equivalent to

(x-\}(x-

30) (x

2) (x

1).

Determination of Multiple Roots. The Method when they are commenIn applying the method, when any divisor of a n which surable.
of Divisors determines multiple roots
is

found to be a root

is

a divisor of the absolute term of the re-

duced polynomial, we must proceed to try whether it is also a it will be a double root of the If it be found to be a root of the next proposed equation.
root of the latter, in which case

reduced polynomial,
so on.

it

will be a triple root of the proposed

and

any degree there exists only r times one multiple root, repeated, it can be found in this way ; for the common measure of /(a?) and /'(#) will then be of the
of

Whenever in an equation

form

(x

a)''"

1
,

and the

coefficients of this could

not be com-

mensurable

if

a were incommensurable.

Determination of Multiple Roots.

205

Multiple roots of equations of the third, fourth, and fifth degrees can be completely determined without the use of the
process of finding the greatest common measure, as will appear from the following observations
:

(1).

The Cubic.

In

this case multiple roots

is

mensurable, since the degree

distinct roots being repeated.
(2).

must be comnot high enough to allow of two

The Biquadratic.

In

this case either the multiple roots

is

a perfect square. For the only form of biquadratic which admits of two distinct roots
are commensurable or the function

being repeated

is

be incommensurable.

The roots of the quadratic may the square of a quadratic. If we find, therefore, that a biquadratic has no commensurable roots, we must try whether it is a per/.

c.

fect square in order to

determine further whether

it

has equal

incommensurable
(3).

roots.

The Qm'ntic.

In

this case, either the multiple roots are

factor multiplied

commensurable, or the function consists of a linear commensurable by the square of a quadratic factor. For, in

order that two distinct roots

may

be repeated, the function must

take one or other of the forms

(*

a) (x

/3) (*

7),

(x

a}\x

/3)

In the
former

latter case the roots

cannot be incommensurable

but the

may correspond to the case of a commensurable factor multiplied by the square of a quadratic whose roots are incommensurable. If then a quintic be found to have no commensurable roots it can have no multiple roots. If it be found to have one commensurable root only, we must examine whether the remaining factor is a perfect square. If it have more than one commensurable root, the multiple roots will be found among

the commensurable roots.

206

Solution of Numerical Equations.

EXAMPLES.
1.

Find

all

the commensurable roots of

2*3 -3L*; 2

+
1,

112*
16.

64

0.

The

roots lie

between the limits

64

The

divisors are 2, 4, 8.

112
8

-31
15

-2
0;
:

120
8 is therefore a root.

-16

Proceed

now with

the reduced equation

-8

-15
1

-2
0;

16
is

is

a root again, and the remaining factor

2x

1.

^,/(*)^
2.

(2*

+!)(*-

8)

Find the commensurable and multiple

a;4

roots of

_x _
s

3o#2

7 6a,

_56 =

0.

The

roots lie

between the limits

6, 12.

(Apply method of Ex. 10, Art. 80).

^m./(*)a(ar + 2)(*-7).
3.

Find the commensurable and multiple roots of

9*4

12* 3

71* 2
2,

- 40* +
5.

16

0.

The

roots lie

between the limits

is

The equation as it stands a commensurable root. To test this

of the coefficient of

found to have no integral root

but

it

may

still

have

we

multiply the roots by 3 in order to get rid

#4

"We find then

144

0.

Limits

6,

15.

We find
(x
z

- 12# +
:

4 to be a double root of this, and the function to be equivalent to 2 The original equation is therefore identical with the fol9) (x + 4)
.

lowing

(X*

- 4* +

1)

(3*

4)

0.

4.

Find the commensurable and multiple

z4

roots of

12* 3

32s2

24s

0.

New fats Method

The
roots lie
1
.

of Approximation.

207

1. The only divisors to be tested are, therefore, equation has no commensurable root. "We proceed to try whether the given function is a perfect square. This can be done by find extracting the square root, or by applying the conditions of Ex. 8, p. 123.

between

- 12 and

4,

2,

We find that the

+ 6x -

We

that

it is

the square of x*

2(cf.

Ex.

1,

p. 161).

Hence the given equation

has two pairs of equal roots, both incommensurable.

5.

Find the commensurable and multiple roots of

f(x)

= x5 -

x*

12*3
4.

8z2

28x + 12

0.

The

limits of the roots are

- 4,
x*

We find that - 3 is a root,

and that there
is

and that the reduced equation

is

& + 8x + 4 = 0,
root.

no other commensurable

case of possible occurrence of multiple roots is, therefore, when this latter function is a perfect square. It is found to be a perfect square, and we have

The only

6.

Find the commensurable and multiple

f(x)

roots of

= x5 -

8z*

+ 22Z3 -

26x~

+ 2lx -18=0.
f(x)

Am.
7.

(x*
:

1) (x

2) (x
:

- 3) 2

The

following equation has only two different roots

find

them

x5

13z4 + 67s 3

- 17U 2 +

2l6x - 108

0.

In general it is obvious that if an integral root' A occurs twice, the last coefficient must contain A2 as a factor, and the second last h if the root occurs three times, A3 must be a factor of the last, A 2 of the second last, and A of the third last coef;

ficient.

The

last coefficient

here

=
3.

22

the required roots

8.

must be 2 and

3 1 nor 1 is a root, Hence, if neither That these are the roots is easily verified.
.

The equation
800z4
:

102# 2

-x+

has equal roots

find all the roots.

it is

In

this

example

convenient to change the roots into their reciprocals before

applying the Method of Divisors.

Ans. f(x)

= (Wx -

3)

(5*

- l)(x +

I)

2.

100. Utewton's Method of Approximation. Having shown how the commensurable roots of equations may be obtained, we proceed to give an account of certain methods of

obtaining approximate values of the incommensurable roots.

The method

of approximation,

commonly

ascribed to Newton,*

which forms the subject of the present

* See Note

Article, is valuable as

at the

end of the volume.

208

Solution of Numerical Equations.

being applicable to numerical equations involving transcendental

as well as those involving algebraical functions only. when applied to the latter class of functions Newton's
for practical purposes, inferior in

form to

Although method is, Homer's, which will

be explained in the following Articles, yet in principle both methods are to a great extent identical.

In
in a

all

supposed to be separated

methods of approximation the root we are seeking is from the other roots, and to be situated
interval between close limits.

known

be a given equation, and suppose a value a to Let/(#) = be known, differing by a small quantity h from a root of the We have then, since a + h is a root of the equation, equation. = or + h) f\a
;

/()+/>) A + -^|/
Neglecting now, since h first, we have
is

'

+ .... =

0.

small, all powers of h higher than the

/(*)+/>) A =

0,

giving, as a first approximation to the root, the value

-78/w
Representing this value by &, and applying the same process a second time, we find as a closer approximation
-

/w
approximation can be carried

By
to

repeating this process the

any

degree of accuracy required.

EXAMPLE.

Find an approximate value

of the positive root of the equation

xz

2z

0.

root lies between 2 and 3 (Ex. 1, Art. 89). Narrowing the limits, the root is found to lie between 2 and 2-2. "We take 2-1 as the quantity represented by a. It

The

cannot

differ

from the true value a

h of the root by more than

0*1.

"We find

Homer's Solution of Numerical Equations.

A
first

209

approximation

is,

therefore,

2-1

-0-00543 =2-0946.

Taking

this as b,

and calculating the fraction

"77777-,

we

obtain

=
for a second approximation
;

2-09455148

and

so on.

The approximation in Newton's method is, in general, rapid. When, however, the root we are seeking is accompanied by an-/*/

other nearly equal to

it,

the fraction

'-7p-

is

not necessarily small,

since the value of either of the nearly equal roots reduces/*^) to case of this kind requires special precaua small quantity.

do not enter into any further discussion of the for practical purposes it may be regarded as since method, entirely superseded by Homer's method, which will now be
tions.

We

Method of Solving; Numerical Equaboth the commensurable and incommethod By mensurable roots can be obtained. The root is evolved figure by figure first the integral part (if any), and then the decimal
tions.
this
:

explained. 101. Homer's

part,

till

the root terminates

of places required if

if it
it

number
and

be commensurable, or to any be incommensurable. The pro-

cess is similar to the

processes of extraction of the square cube root, which are, indeed, only particular cases of the

known

general solution
equations.

by

the present

method

of quadratic

and cubic
is

The main

principle involved in

Homer's method

the suc-

cessive diminution of the roots of the given equation by known The great adquantities, in the manner explained in Art. 33.

vantage of the method

are

is,

that the successive transformations

exhibited in a compact arithmetical form, and the root obtained by one continuous process correct to any number of
places of decimals required.
*

This principle of the diminution of the roots will be illusIn the trated in the present Article by some simple examples.

210

Solution of Numerical Equations.

following Articles

we

shall proceed to certain

considerations

which tend to

facilitate the practical application of the

method.

EXAMPLES.
1.

Find the

positive root of the equation

2*3

85z2

85x

87

0.

Jirst figure of the root.

equation is proposed for solution, is to find the This can usually he done by a few trials although in certain cases the methods of separation of the roots explained in Chap. IX. may have
first step,
;

The

when any numerical

to

it is

he employed. In the present example there can he only one positive root and found by trial to lie between 40 and 50. Thus the first figure of the root is 4.
;

between

"We now diminish the roots by 40. The transformed equation will have one root and 10. It is found by trial to lie between 3 and 4. We now diminish
;

so that the roots of the proposed equathe roots of the transformed equation by 3 The second transformed equation will have one root tion will be diminished by 43.

between
that
its

and

1.

On

absolute term

is

diminishing the roots of this latter equation by *5, we find reduced to zero, i. e. the diminution of the roots of the pro-

posed equation by 43*5 reduces its absolute term to zero. We conclude that 43-5 is a root of the given equation. The series of arithmetical operations is represented as
follows
:

-85
80

(43-5

Examples.

211

is the equation whose roots are each less by 40 than the roots of the given equation, and whose positive root is found to lie between 3 and 4. If the second transformed equation had not an exact root -5 but one, we shall suppose, between '5 and -6, the
;

first

three figures of the root of the proposed equation

would be 43-5

and

to find

the next figure

we should proceed

to a further transformation, diminishing the roots

by

-6
2.

and so on.

Find the positive root of the equation

4*3

13* 3

- 3Lr -

275

0.

We first write down the arithmetical work,

on
it
:

and proceed to make certain observations

13

24

212
the process
is

Solution of Numerical Equations.

continued as befpre.

To
all

illustrate this

we

repeat the above operation,

omitting the decimal points.

In

subsequent examples this simplification will be

adopted

- 13
24

(6-25

Principle of the Trial-divisor.

102.

213
have seen in

Principle of the Trial-divisor.

for

We

Art. 100 that

of a
-f
//

when an equation is transformed by the substitution X, a being a number differing from the true root by
is

a quantity h small in proportion to a,

cal value of h

an approximate numeriobtained by dividing f(a) by /'(). Now the

being

successive transformed equations in Horner's process are the

results of transformations of this kind, the last coefficient
.

and the second last f(a) (see Art. 33). Hence, after two or three steps have been completed, so that the part of the root
remaining bears a small ratio to the part already evolved, we may expect to be furnished with two or three more figures of the
root correctly by mere division of the last coefficient of the final transformed equation.
fore, if

by the second last -We might there-

we

Newton's

pleased, at any stage of Homer's operations, apply method to get a further approximation to the root. In
is

Horner's method this principle

employed

to suggest the

next

following figure of the root after the figures already obtained. The second last coefficient of each transformed equation is called the trial-divisor. Thus, in the second example of the last Article,

the

number 5
In
this

2690800
root
is

is correctly suggested by the trial-divisor example, indeed, the second figure of the

correctly suggested

by the

trial-divisor of the first trans-

formed equation, although, in general, this is not the case. In practice the student will have to estimate the probable effect of
the leading coefficients of the transformed equation he will find, however, that the influence of these terms becomes less and less
;

as the evolution of the root proceeds.

EXAMPLES.
1
.

Find the positive root of the equation

xt

100

to foir- decimal places.

It is easily

seen that the root

lies

between 4 and
it
:

5.

We

write

down

the work,

and proceei

to

make

observations on

214

Solution of Numerical Equations.

i

- 100
84

(4-2644

20
5

21

36

Principle of the Trial-Divisor.

and
this will

215
In

show

itself in

the

work by the change

is 4,

of sign in the absolute term.

the above work

it is

evident, without performing the fifth transformation, that the

so that the correct root to four decimal places

-ponding figure of the root

2.

The equation
1

x*

4a?

- 4*2 -

11*

-I-

has one root between

and 2

-\find its value correct to four decimal places.

-/4

216
We
The
has

Solution of Numerical Equations.

see without completing the fifth transformation that 9 is the next figure of

the root.

The

root

is,

therefore, 1-6369 correct to four decimal places.

effective after the second transformation, suggesting

trial-divisor

becomes

correctly the
its last

number 3, and all subsequent numbers. two terms negative. We may expect,

The

first

transformed equation

therefore, that the influence of

the preceding coefficients is greater than that of the trial- divisor, as in fact is here the case. The number 6, the second figure of the root, must be found by substitution.

We hare to

determine what
x*

is

the situation between

and 10 of the root of

0.

the equation

80* 3

1400*2

3000-r

60000

few

trials

show

that 6 gives a negative,

;

root lies between 6 and 7

and 6

is

the

number

and 7 a positive result. Hence the In the of which we are in search.

subsequent trials we take those greatest numbers 3, 6, 9, in succession, which allow the absolute term to retain its negative sign. In the first transformation, diminishing the roots by 1, there is a change of sign in the absolute term. The meaning of this
is,

that

we have
;

subsequent transformations, so long as we keep below the root, the sign of the absolute term must be the same as the sign resulting from the Bubstitution of 1. This supposes of course that no
root lies

result, 4

and

passed over a root lying between 6. In gives a negative result,

and
all

1,

for

gives a positive

already

between 1 and that of which we are in search. This supposition we have made in the statement of the question. In fact the proposed equation can have only two positive roots one of them lies between and 1, and therefore only
;

one between

and

2.

When two
i.e.

roots

exist

between the

limits

employed in Homer's method,

the equation has a pair of roots nearly equal, certain precautions must be observed which will form the subject of a subsequent Article.
3.

when

Find the root of the preceding equation between

10.

and

to four decimal places.

Commence by multiplying by

The

coefficients are

then
;

1, 40,

400,

11000, 40000

the trial-divisor becomes effective at once in consequence of the comparative smallness of the leading coefficients. The positive sign, of the absolute term must be preserved throughout.
4.

Am.

-3373

Find

to three places of decimals the root situated

between 9 and 10 of the

equation
#4

_ 3^2 + 753; __
z 3 .]

[Supply the zero

coefficient of

Am.

9-886.

In the examples hitherto considered the root has been found few decimal places only. We proceed now to explain a method by which, after three or four places of decimals have
to a

been evolved as above, several more may be correctly obtained with great facility by a contracted process.

Contraction of Homer's Process.

103.

217

Contraction of Homer's Process. In the ordinary process of contracted Division, when the given figures are
appending ciphers to the successive divioff cut 1s, figures successively from the right of the itself becomes exhausted after a numthe divisor so that divisor,
iiisted,

in place of

we

ber of steps depending on the

number

resulting quotient will differ from ilgure only, or at most in the last two figures. contracted method the principle is the same.

The

of figures it contains. the true quotient in the

In Horner's
retain those

We

figures only which are

effective in contributing to the result to

the degree of approximation desired.

When

the contracted

process commences, in place of appending ciphers to the successive coefficients of the transformed equation in the way before

explained,
ficient

we

cut off one figure from the right of the last coeflast coefficient
;

but one, two from the right of the

but two,

three from the right of the last coefficient but three and so on. The effect of this is to retain in then- proper places the important figures in the work, and to banish altogether those which
are of
little

importance.
will

The student

do well to compare the

first

transformation

in the first of the following examples with the corresponding step in the second example of the last He will Article, where the transformation is exhibited in full.

by the contracted process

the leading figures (those which are most important in contributing to the result) coincide in both cases, and retain their relative places while the figures of little im-

then observe

how

portance are entirely dispensed with. In addition to the contraction now explained, other abbreviations of Horner's process are sometimes recommended ; but as the advantage to be derived from them is small, and as they
increase the chances of error,

give any account of them.

of so

we do not think it necessary to The contraction here explained is

much importance
it.

method

of approximation that

in the practical application of Horner's no account of this method is

complete without

218

Solution of Numerical Equations.

EXAMPLES.

1.

Find the root between

and 2 of the equation in Ex. 2 of the

last Article

correct to seven or eight decimal places.

Assuming the
work stands

result of the

Example

referred to,

we

shall

tracted process after the third transformation has been completed.

as follows
:

commence the conThe subsequent

2516578H 18936
3156
6

17549439

(1-636913575

15213090

2535515
18972

2336349

2301597

3162

2554487 285
255733
285
256018,

34752
25601

-9151
7680

1471

1280

191

179
12

Here the

effect of the first cutting off of figures,

coefficient, 14

from the third

last,

and 052 from the fourth

namely, 8 from the second last last, is to banish alto-

"We proceed to diminish the roots by gether the first coefficient of the biquadratic. 6 as if the coefficients 1, 3150, 2516578, - 17549439 which are left were those of a
cubic equation. In multiplying by the corresponding figure of the root the figures cut off should be multiplied mentally, and account taken of the number to be carried, just as in contracted division.

After the diminution by 6 has been completed, we cut off again in the transformed cubic 7 from the last coefficient but one, 68 from the last but two, and the first
coefficient disappears altogether.

with the

coefficients 31,
is

process of cutting off

The work then proceeds as if we were dealing 255448, -2336349 of a quadratic. The effect of the next The subto banish altogether the leading coefficient 31.

sequent work coincides with tbat of contracted division. "When the operation terminates, the number of decimals in the quotient may be depended on up to the last two or three figures. The extent to which the evolution of the root must be
carried before the contracted process
places required
;

for after the contraction

tion to the figures already evolved,

trial -divisor, less one.

commenced depends on the number of decimal commences we shall be furnished, in addiwith as many more as there are figures in the
is

2.

Find

to seven or eight decimal places the root of the equation

&which
lies

12x + 7

between 2 and

3.

Examples.

219

This equation can have only two positive roots : one lies between and 1, and the other between 2 and 3. For the evolution of the latter we have the following
:

220

Solution of Numerical Equations.

104. Application of Homer's Method to Cases where Roots are nearly Equal. We have seen in Art. 100 that the method of approximation there explained fails when the pro-

posed equation has two roots nearly equal. Examples of this nature are those which present most difficulties, both in their
analysis (see Ex. 7, Art. 91)

and in

their solution.

By Homer's

possible, with very little more labour than is necesin So other cases, to effect the solution of such equations. sary
it is

method

long as the leading figures of the two roots are the same certain precautions must be observed, which will be illustrated by the After the two roots have been separated, following examples.
the subsequent calculation proceeds for each root separately, just as in the examples of the previous Articles. It is evident, from
the explanation of the trial-divisor given in Art. 102, that for the same reason as that which explains the failure of Newton's

method in the case under consideration (see Art. 100), it will not become effective till the first or second stage after the roots
have been separated.
EXAMPLES.
1.

The equation
tf
1

1x

=
;

has two roots between

cimal places.

and 2

(see

Ex.

2,

Art. 89)

find each of

them

to eight de-

Diminishing the roots by

roots are multiplied

1,

we

find that the transformed equation (after its

by

10), viz.

#3

30^2

4003;

1000

0,

and we proceed with each separately in the manner already explained. If the roots were not separated at this stage, we should find the leading figure common to the two, and,

must have two roots between and 10. 3 and 4, and the other between 6 and 7.

We
The

find that these roots

lie,

one between

roots are

now

separated,

having diminished the roots by it, find in what intervals the roots of the resulting equation were situated and so on.
;

Am.

1-35689584,

1-69202147.

2.

Find the two

roots of the equation

x*

- 49z2 + 658# -

1379

which

lie

between 20 and 30.

roots to seven places

"We shall exhibit the complete work of approximation to the smaller of the two and then make certain observations which will be a guide to
;

the student in

all cases of

the kind.

-49
20

222

Solution of Numerical Equations.

us to find by mere substitution the interval within -which the two roots lie. If we diminish the roots of * 3 + 11#2 - 102* + 181 = by 4, the resulting equation is xz -f 23*2 + 34* + 13=0, which has no change of sign. Hence the two roots must

and 4. If we diminish its roots by 3, the resulting equation (as in between the above work) has the same number of changes of sign as the equation itself. Hence the two roots lie between 3 and 4. They are, therefore, not yet separated
lie
;

and we proceed to diminish by

x*
is

3.

The next transformed equation

900*

+ 200*2 -

1000

=
3
:

found in the same way to have both

its

roots

between 2 and

the diminution

by

2 leaving two changes of sign in the coefficients of the transformed equation (as in So far, then, the the above work) and the diminution by 3 giving all positive signs.
,

two

roots agree in their first three figures, viz. 23-2.

We diminish again by 2.
:

The
1

3 resulting equation *

+ 2060*2 - 8800* + 1261000=

has one root only between

giving a positive, and 2 a negative result its other root lies between 2 and 3 3 giving a positive result. The roots are now separated. proceed, as in the above work, to approximate to the lesser root, by diminishing the roots of this

and 2

We

diminish by 2 the roots of the same equation, taking care that in the subsequent operations the negative sign, to which the preThe second viously positive sign of the absolute term now changes, is preserved.
root,

the equation by 1 mate to the greater

;

trial divisor

becoming

effective at the next step.

To

approxi-

we must

root will be found to be 23-2295212.

may

So long as the two roots remain together, a guide to the proper figure of the root be obtained by dividing twice the last coefficient by the second last, or the se-

cond last by twice the third last. The reason of this is, that the proposed equation approximates now to the quadratic formed by the last three terms in each transformed
equation, just as in previous cases, and in Newton's method it approximated to the simple equation formed by the last two terms, this quadratic having the two nearly z = equal roots for its roots ; and when the two roots of the equation az + bx + c
are nearly equal, either of

them is given approximately by

is

or

Thus, in the
.

above example, the number 3

900 In this way we can generally, at the first attempt, find the two integers between which the pair of roots lies. We shall have, also, an indication of the separation of
the roots

suggested by

102

and the number 2 by

by observing when the numbers suggested

become
different, i.e.

in this

way by

the last three

.

coefficients
3.

when

suggests a different

number from

Calculate to three decimal places each of the roots lying between 4 and 6 of
**

the equation

8* 3

70*2

144*

936

0.

Ans. 4-242; 4-246.

4.

Find the two

roots

between 2 and 3 of the equation

64* 3

592* 2

-f

1649*

1445

0.

Ans. The roots are both

2-125.

Lagrange
Here we

Method of Approximation.

223

find that the two roots are not separated at the third decimal place. diminish by 5 the absolute term vanishes, showing that 2-125 is a root; and proceeding with this diminution the second last coefficient also vanishes. Hence

When we
2-12') is

a double root.

When
roots,

an equation contains more than two nearly equal they can all be found by Homer's process in a manner

similar to that

now

explained.

Such

cases are, however, of

rare occurrence in practice.

principles already laid down will be a sufficient guide to the student in all cases of the kind.

The

105.

Lagrange's Method of Approximation.

Lagrange

has given a method of expressing the root of a numerical equaAs this method is, for tion in the form of a continued fraction.
to that of Horner, practical purposes, much inferior content ourselves with a brief account of it.

we

shall

Let the equation f(x) = have one root, and only one root, between the two consecutive integers a and a + 1. Substitute
a + - for x in the proposed equation.

The transformed equation

in y has one positive root.

lie

between the integers

if

Let this be determined by trial to and b + 1. Transform the equation

b -f -.

in

by the

substitution
s is

y =

The

positive root of the

equation in

tinuing this process,

found by trial to lie between c and c+1. Conan approximation to the root is obtained in
:

the form of a continued fraction, as follows

1

a +

+ 1

EXAMPLES.
1
.

Find in the form of a continued fraction the positive root of the equation X3

2X

o.

The

root lies

between 2 and

3.

To make

the transformation x

-f

-,

we

first

employ the process of Art.

33,

diminishing the roots by 2. "We then find the equation whose roots are the reciprocals of the roots of the transformed.

224
The

Solution of Numerical Equations.

equation in y
is

in this

way found

to be

This has a root between 10 and 11.

Make now

the substitution
in z
is

y = 10 +

-.

The equation

61z 3

94z2

20z

=
+

0.

This has a root between

and

2.

Take

=
-

-.

The equation

in

is

54w 3

+ 25w2 ;

89?<

61

0,

which has a

root

between

and 2

and so on.

We have, therefore,

the following expression for the root

w+I
2.

Find in the form of a continued fraction the positive root of

x*

fa

13

0.

Ans, 3

106.

Numerical Solution of the Biquadratic.

It

is

proper, before closing the subject of the solution of numerical equations, to illustrate the practical uses which may be made of

the methods of solution of Chap. YI. Although, as before observed, the numerical solution of equations is in general best effected by the methods of the present Chapter, there are certain

which it is convenient to employ the methods of for the resolution of the biquadratic. VI. When a biChap. leads to a has a comcubic which quadratic equation reducing
cases in

mensurable root,
of the

be readily found, and the solution "We proceed to solve a few completed. of this Descartes' method (Art. 64), which kind, examples using will usually be found the most convenient in practice.
this root can

biquadratic

Examples.

225

EXAMPLES.
1.

Resolve the quartic

2

22z

into quadratic factors.

Making the assumption

of Art. 64,

we
3,

easily obtain

=
Also

3,

q'

+ 4pp =
'

pq'

+ p'q=H,

qq'

=-

6.

<

--pp =

_ q (

+ q '-i)

and, calculating

/ and

/, the equation for

<f>

is

VMultiplying the roots by

4,

111

225

*-8if 4<j>

0.

we

have,

t,

PBy
</>

111*- 450 =

0.

the

Method

of Divisors this is easily found to have a root

hence

= --,

giving

pp' =

2,

+ q'=-

5.

From

these,

combined with the preceding equations,

we

get

p = -2,

/ = -!,

q=l,

q'=-6.

"When the values of q and q are found, the equation giving the value ofpq'+p'q determines which value of q goes with p, and which with p', in the quadratic The quartic is resolved, therefore, into the factors factors.
(y?

_ X+
of

1) (a;2

_ 2# -

6).

By means
factors in

of the other

two values
;

^>

we can

resolve the quartic into quadratic

two other ways

or

we

can do the same thing by solving the two qua-

dratics already obtained.

2.

Resolve into factors the quartic

/(*)

x*

- S*3 - 12z2 + 60* +

63.

The equation

for

q>

is
3
4<j>

5.

195<J>

475 =

0,

which

is

found to have a root

Am.

f(x)

(x

- 2z -

3) (**

- 6* -

21).

226
3.

Solution of Numerical Equations.

Eesolve into factors

f(x)

z*

- 17s2 - 20s -

6.

The reducing

cubic

is

found to be

_
12
or,

216

multiplying the roots by

6,

4* 3

651*

+ 3185 =

0.

This has a root

hence

<

-T.

Am.
4.

f(x)

E=

(a?

+ 4s +

2) (x*

- 4s -

3).

Eesolve into factors

f(x)

6a; 3

9*2

66*

22.

The reducing

cubic

is

hence

^ = -

-.

^#w.
5.

/().=

(^

11)

fa +

2).

Eesolve into factors

f(x)

= x*-

8x3

1\x>

'26x

+
(x>

14.

Am.
6.

f(x)

2x +

2) (2?

6z

7).

Eesolve into factors

x*

+ 12x +
(x*

3.

Am.
7.

x^/6

\/6) (x

+ x \/Q +

\/6).

Find the quadratic

factors of
*

8*3

12s2

84s

-63 =

0,

and solve the equation completely

Ans. {x*

(see

Ex.

18, p. 34).

-2x(2 + V7) +

Vf) {s

- 2s

(2

\/7)

- 3V7}.

Examples

227

EXAMPLES.
1.

Find the positive root of

s3_ 6* -13 =
2.

0.

Ant. 3-176814393.

Find the positive root of

x3

2* - 5

=
Ans. 2-094551483,

correct to eight or nine places.

3.

The equation
2s3

650-Sz2
:

+ 5* -

1627

has a root between 300 and 400

find

it.

Ans. Commensurable
4.

root, 325-4.

Find the root between 20 and 30 of the equation 4s3

180 *3

+ 1896* and

457

0.

Ans. 28-52127738
5. Find to six places the root between 2

3 of the equation

1370

0.

Ans. 2-557351.
6.

Find to six places the root between 2 and 3 of the equation

12*
7.

0.

Ans. 2-858083.

Find the positive root of the equation

s3

2*2

23*

70

=
Ans. 5-13457872528.

correct to about ten decimal places.

8. 9.

Find the cube root of 673373097125. Find the Find

all

Ans. 8765.

fifth root of

537824.

Ans.

14.

10.

the roots of the cubic equation

*3

3*

0.

The equation

&+*

0,

of

Ex.

7, p.

100, reduces to this.

Ans.
Note.

1-87938,

-34729,

1-53209.

Tho

smaller positive root furnishes the solution of the problem

is

To

divide a hemisphere whose radius

to the base.
11
.

unity into

two equal parts by a plane

parallel

Find

all

the roots of the cubic

*3

+ *2 ~

2*

0.

(See Ex.

1, p.

100.)

Ans. -

1-80194,

-44504,

1-24698.

Q2

228
12.

Solution
Find
to five decimal

*'

Numerical Equations.
the negative root between

and

(see

Ex.

3,

p. 100) of the

equation

Am. 13. Solve the equation

-28463.

x3

315z 2

- 19684# + 2977260 =

0.

found

between 70 and 80. By Homer's process it is depressed equation furnishes two roots, which, increased by 78, are the other roots of the cubic.

"We here
to

find that a root exists

be 78.

The

Am.
14.

78,

347,

110.

Find the two

real roots of the equation

z*

-11727* + 40385 =

0.

Am.
This equation
Viscous Spheroid,
is

3-45592,

21-43067.

and on

given by Mr. G. H. Darwin in a Paper On the Precession of a the Remote History of the Earth. Phil. Trans., Part ii.,

1879, p. 508.
tion for
15.

The

roots are

"the two values

of the cube root of the earth's rotaas a rigid

which the earth and moon move round

Find
all

body."

the roots of the cubic equation

20z3 - 24s 2

0.

Am. -

0-31469,

0*44603,

1-06865.

This equation occurs in the solution by Professor Ball of a problem of Professor Townsend's in the Educational Times of Dec. 1878, to determine the deflection of a beam uniformly loaded and supported at its two ends and points of trisection.
16.

Find the positive root of the equation

*2

9z

10

0.

Am.

0-85906.

The equations of this and the following example occur in the investigation of to beams supported by props. questions relative
17.

Find the

positive root of the equation

lx^

20o; 3

3*2

16*

0.

Am.
18. Find to ten decimal places the positive root of the equation

0-91336.

x*

12z 4

+ 59s3 + 150*2 + 201* -

207

0.

Ans. -6386058033.

CHAPTER
107.

XI.

DETERMINANTS.

Elementary Notions and Definitions.

an important

This Chapter
class of func-

will be occupied with a discussion of

tions which constantly present themselves in analysis.

These

much

functions possess remarkable properties, by a knowledge of which simplification may be introduced into many mathematical
operations.

The

function

a^ +

a z b lf of the four quantities

is

obtained by assigning to a and

1, 2,

b,

written in alphabetical order,

the suffixes
of the

and
2
;

2, 1,

numbers

1,

corresponding to the two permutations and adding the two products so formed,

Similarly the function

fli& 2
3

dibaCz

chb^Ci

dzbiCs

+ a 3 biCz + a3 b 2 Ci,

(1)

of the nine quantities

a3
is

c3

obtained by adding all the products by assigning to the letters (retained in

suffixes corresponding to all the
1, 2, 3.

dbc which can be

formed

their alphabetical order)

permutations of the numbers expression might be represented by (abc), or any other convenient notation, from which all the terms could

The whole

be written down.

230

Determinants.

The notation (abed) might be employed to represent a similar function of the 16 quantities a i9 b h c^ d^ # 2 &c. consisting of 24 terms, which can all be written down by the aid of the 24 permu, ;

tations of the

numbers

1, 2, 3, 4.
. .
.

write And, in general, taking n letters a, b, c, Z, we can down a similar function consisting of n (n - 1) (n - 2) .... 3. 2. 1 terms, this being the number of permutations of the first n numbers 1, 2, 3 ... n.

Now

the functions above referred

to,

which are of such

frequent occurrence in mathematical analysis, differ from those of the 1. 2. 3 ... n just described in one respect only, namely
:

(which
tive,

is

an even number) terms, half are affected with

posi-

and half with negative

signs, instead of being all positive,

as in the examples just given. shall now give some instances of these latter functions.

We

They

variables from linear equations. eliminated from the equations

occur most frequently as the result of elimination of the If, for example, x and y be

a& +

%=

0,
0,

a 2x + b^y =

the result

is

ajb*

- a^ =

0. y, %

Again, the result of eliminating x

a^x
a-iX

from the equations

+ hy + C& =

0,
0,

b zy

c<&

is,

as the student will readily perceive

by

solving from two of

the equations and substituting in the third,

ftibzCa

aib 3 c2

azbrf!

- a z biC3 + a z biC2 - a 3 bCi =

(1)

(2)

and
tive.

this function differs

its

three of

above written only in having terms negative, instead of having the six terms posi-

from

Elementary Notions and Definitions.

231

Similarly the elimination of four variables from four linear equations gives rise to a function of the sixteen quantities
flit

^1J C U dl)

f*Z>

^3,

&G-9
(abed) only

which

differs

from the function above represented by

in having twelve of its terms negative.

Expressions of the kind here described are called Determinants* The notation by which they are usually represented was
first

employed by Cauchy, and possesses many advantages in the treatment of these expressions. The quantities of which the function consists are arranged in a square between two vertical

lines.

For example, the notation

represents the determinant Similarly, the expression on the left-hand side of equation (2) is represented by the notation

And,
i,

in
.

general,
#2, #2,

the determinant of the n* quantities

bi,

/i,

&c., is represented
ftl

by

tfl

(3)

By taking
to

the n letters in alphabetical order, and assigning

2)
.
.

them

suffixes corresponding to the n(n- - l)(w

3. 2. 1

permutations of the

numbers

1, 2, 3,

...

n, all

the terms of the

232

Determinants.

determinant can be written down.

receive positive

the rule will

Half of the terms must and half negative signs. In the next Article he explained by which the positive and negative
ai, b h GI,

terms are distinguished.

The

individual letters
is

a2

&c., of

determinant

composed, are called

constituents,

which a and by some

writers elements.

such as a, b ly Ci, A, arranged form a row of the determinant and any series such horizontally, as a\, 2> #3, a n arranged vertically, form a column.

Any

series of constituents

The term
column
108.

line will

sometimes be used to express a row or

evident from

indifferently.

Rule with regard

to Signs.

It

is

the preceding Article that each term of the determinant will,

since
it contains all the letters, contain one constituent (and only from one) every column and will also, since the suffixes in each term comprise all the numbers, contain one constituent (and only
;

one) from every row. may thus regard the square array of Art. 107 as the (3) symbolical representation of a function consisting in general of

We

n(n-l)(n 2) ... 3. 2. 1 terms, comprising all possible products which can be formed by taking one constituent, and one only, from each row and one constituent, and one only, from each column, All that is required to give perfect
;

definiteness to the function is to fix the sign to be attached to any For this purpose the following two rules are particular term.
to be observed
(1).
:

in the diagonal

The term aib 2 c3 Informed by the constituents situated drawn from the left-hand top corner to the right.
. .

.hand bottom corner,

is

positive.

This

is

called the leading or principal term.

In
;

it

the suffixes
it

and

letters

both occur in their natural order

is

and from

the

sign of any other term

(2)

obtained by means of the following rule.

Any

interchange of two suffixes, the letters retaining their

order, alters the sign of the term.

Rule

ivith

regard

to

Signs.
:

233

This rule may be otherwise expressed thus

of two term.

Any

interchange

tetter*, the suffixes retaining their order, alters the sign

of a

For

if

two

letters

be interchanged, and the two corre-

sponding constituents interchanged, the process is equivalent to an interchange of suffixes. If, for example, in a^c^d^ the
letters b

and

to aib 5 cA de?,

and

be interchanged, we get a^e^c^dj)-^ which is equal this is derived from the given term by an in5.

terchange of the suffixes 2 and

In applying

this rule it is evident that

an even number of

interchanges will not alter the sign of a term,

and that an odd

number

will.

EXAMPLES.
1.

What
?

sign

is

to be attached to the

term a^b^c^d^e^ in the determinant of the

5th order

Here,

The question is, How many interchanges will change the order 12345 into 34251 ? when 3 is interchanged with 2, and afterwards with 1, it comes into the lead-

ing place, the order becoming 31245. Again, the interchange in 31245 of 4 with The interchange of 2 with 1 2, and afterwards with 1, presents the order 34125.
gives the order 34215
;

and

finally the interchange of 5

;

with

gives the required

order 34251.
sign

Thus there

are in all six interchanges

and therefore the required

is positive.

The general mode

figure

of proceeding

may

plainly be stated as follows

Take the
place in the

which stands

first

in the required order,

and move

it

from

its

natural order 1234 ... into the leading place, counting one displacement for each Take then the figure which stands second in the required order, figure passed over.

and move
the

it

from

its

place in the natural order into the second place

and so on. If
;

number

of displacements in this process be even, the sign is positive

is

if it

be

odd, the sign

negative.
is

2. What sign the 7th order ?

to be attached to the

term

l o.-^'b-

c^d) e\f^gi in the determinant of

Here two displacements bring 3 to the leading place five displacements then bring 7 to the second place four then bring 6 to the third place three then bring 5 to the fourth place the figure 1 is in its place and finally, one displacement Thus there are in all fifteen displacements and the brings 4 into the sixth place.
; ; ; ; ; ;

required sign
3.

is

therefore negative.
all

Write down

the terms of the determinant

f>\

234
The

Determinants.
six permutations of suffixes in

which the
1342,

figure 1 occurs

first

are

1234,

1243,

1324,

1423,

1432.

The
Rule

six corresponding

terms are, as the student will easily see by applying the

(2),

as in the previous examples,

The other

2, 3, 4, respectively,

eighteen terms, corresponding to the permutations in which the figures stand first, are as follows
:

a^

It will be observed here that the

of negative terms.

number of positive terms is equal to the number The same must be true in general for, corresponding to any
;

positive term there exists a negative term obtained

by simply interchanging the last

together over any

two

suffixes.

4.

Show
if

that if

any two adjacent

is

figures are

moved
is

number

m of figures,
For
over
5.

the sign

unaltered.

they be moved separately, the whole process

equivalent to a

movement
viz.,

2m

figures.

anbn-i cn-"i

Determine the sign to be attached to the second diagonal term, ... ^2 h, in the determinant of the n th order.
to
is

Here the number of displacements required

required order
plainly

change the natural order to the

(n

1)

2)

(n
n

3)

2 +

^~

(*-!)

Hence the required

sign

is

2
(

1)

109. In the Propositions of the present and following Articles are contained the most important elementary properties of determinants which, by the aid of Cauchy's notation above
described, render the
practical advantage.

employment
rows, or

of these functions of such

columns, of a determinant be interchanged, the sign of the determinant i$ changed. This follows at once from the mode of formation (Eule (2),
I.

PROP.

If any two

any

tiuo

Art. 108), for an interchange of two rows

interchange of is the same as

is the same as an two suffixes, and an interchange of two columns an interchange of two letters so that in either
;

case the sign of every term of the determinant

is

changed.

Propositions.

235

By

any term

aid of this proposition the rule for obtaining the sign of may be stated in a form which is often more conve-

It will nient for practical purposes than that already given. readily be perceived that the general mode of procedure ex-

plained in Ex. 1, Art. 108, is equivalent to the following: Briny, by movements of rows (or columns), the constituents of the term whose sign is required into the position of the leading diagonal.

The sign of

the

term will be positive or negative according as the

is

number of displacement*

even or odd.

EXAMPLE.

What

sign

is to

be attached to the term \ftnx in the determinant

a
b
c

x
j

a
I

7
in

y
z

n
v

A.

/j.

Here a movement
brings
A.

of the fourth

into the leading place.

row over three rows (i.e. three displacements} One displacement of the original second row up-

And one further ft into the required place in the diagonal term. displacement of the original third row upwards effects the required arrangement, bringing xftnx into the diagonal place. Thus the number of displacements being
wards brings
odd, the required sign
is

negative.

110. PROP. II.

When,
I.,

in

any determinant, two

rotes or two

columns are
For,

identical, the

determinant vanishes.

by Prop.

the interchange of these two lines ought

to change the sign of the determinant A ; but the interchange of two identical rows or columns cannot alter the determinant in

any way
rotes be

hence

= - A, or

A =

0.

111. PROP. III.

th<-

The mine of a determinant is not altered written as columns, and the columns as row's.

if

For all the terms, formed by taking one constituent from each row and one from each column, are plainly the same in value in both cases the principal term is identically the same
; ;

and

to determine the sign of

any other term (by Prop.

I.)

the

236
number

Determinants.

of displacements of rows necessary to bring it into the leading diagonal in the first case is the same as the number of displacements of columns necessary in the second case.

EXAMPLE.
b\

a\

az
bz
Cz

(iz

dz

bi
C\

bz
Cz

#4

fa

di

dz

dz

Here the sign of any term, e.g. az bc\ ^3, is the same in both determinants. For three displacements of rows are required to bring this term into the leading position in the first determinant ; and the same number of displacements of .columns is required to bring the same constituents into the leading position in the second determinant.

112. PROP. IY.

If every

constituent

many

line be

multiplied

by the same factor, the determinant is multiplied by that factor. For every term of the determinant must contain one, and

only one, constituent from any row or any column.

Cor. 1. If the constituents in

same

factor

any line differ only by the from the constituents in any parallel line, the deany line be For this is

terminant vanishes.
Cor. 2. If the signs of all the constituents in changed, the sign of the determinant is changed.

equivalent to multiplying

by the

factor

1.

EXAMPLES.
a\
Jcaz

b\

c\

dz
(tz

bz
bz

Cz
Cz

ai
2.

Mai

az

01

ai

az

01
71

M0l 02
inyi

01
71

01
7'

02 72

72

Examples.
3.

237
:

Show

that the following determinant vanishes

3152 2573 8914

6

15

21

When
4.

the constituents of the last

TOW

are divided

by

3,

they hecome identical

with those of the second row.

Prove the identity

be

a
b

az
*2
c2

ca

ab

tively.

Represent the first determinant by A, and multiply the rows by We have then
abc abc
a"
2

b, c,

respec-

a3

abc abc

b* c2

&
c3

and, dividing the first column by abc, the result follows. 5. Prove the identity

238
It
is

Determinants.
evident that a similar process
all

determinant to one in which

may be employed in general to reduce any the constituents of any selected row or column shall
to one in

be units.
8.

Reduce the following determinant

:

which the

first

row

shall consist

of units

Development of Determinants.
It is evident that a similar

239

proof shows in general that the value of the determi,

.
.

nant of this form, constituted by the n quantities o, A, is the product of the 7 -w (/i - 1) differences which can be formed with these n quantities.
.

113.

Minor Determinants.

Definitions.

When

in a

determinant any number of rows, and the same number of columns, are suppressed, the determinant formed by the remain-

ing constituents (maintaining their relative positions) minor determinant.

If one

is

called a

row and one column only be suppressed, the correIf two rows and two is called .a first minor. minor sponding columns be suppressed, the minor is called a second minor ; and The suppressed rows and columns have common conso on.
stituents
is

forming a determinant

said to be complementary to this determinant.

and the minor which remains The minor

complementary first minor, and

to the leading constituent a l is called the leading

its leading first minor again is the leading second minor of the original determinant. It is usual to denote a determinant in general by A. We

by A a the first minor obtained by suppressing in A and column which contain any constituent a by A a ^ the row the second minor obtained by suppressing the two rows and two columns which contain a and /3 and so on. Thus A 0l represhall denote
;
,

sents the leading first minor,

and

A ai ,j,

or

A a> ,6

the leading

second minor.

The determinant
is

A, formed

by the
.

constituents a ly b l9 c ly &c.,

often denoted for brevity by placing the leading term within A = (a l b z c-A The notation brackets, as follows 4).
:
.

also used to represent A ; this expressing its constitution as consisting of the sum of a number of terms (with their proper signs attached) formed by taking all possible per-

b> c 3

ln is

mutations of the n
114.
of

suffixes.

Development of Determinants.
it

Since every term

any determinant contains one, and only

one, constituent

from

each row and from each column,

follows that

is

a linear and

function of the constituent* of any

one.

row or any one

240
column.

Determinants.

Thus we may write

or,

again,

A = a\Ai + dzAz + a- A$ + &c. A = b.B, + bzBz + b& + &c. A = a^Ai -f b B + c^Ci + &c. A = a A + bzB + CzCz + &c.
A
l
l

The student, on referring to Ex. 3, Art. 108, will observe that the determinant of the fourth order there written at length
is

constituted in the
c2

way

here described, namely

=0*1

We proceed
the form

to

show that in the general

cizAz

case, writing

in

A = aiAi + azAz +
the coefficients Ai,

+ an A n

A A
2,

3,

&c., are determinants of the order

n-l.
In effecting
referred to
i

all

suppose first 1 to
;

the permutations of the suffixes 1, 2, 3. ...;?, remain in the leading place, as in the example
obtain 1
.

we then
and

(n

1)

terms which have

as a factor,

aiAi =

bzc-i

4
.

hence
bz
GZ
. .

lz

= 2

c3

and

this determinant is the

!,

minor corresponding

to the consti-

tuent

or

AI = A 0j

To find the value of A z we bring 2 into the leading place one by displacement of rows. This changes the sign of A, so that we obtain A 2 = - A rta i. e. AZ = the minor corresponding to
, ;

Again, bringing a A to the leading = A two and so on. place by displacements, we have A*
3

with

its

sign changed.

ff

Developmen t of Determinants.
Thus we conclude
that, in general,

24 1

A=
Similarly,

fli

rtl

a2

fla

A 04 +

&c.

we can expand A
any row.
i

in terms of the constituents of

any

other column, or

For example,
Ai x +
Ci

A=

ffl

fli

A Cl -

&c.

If it be required to obtain the proper sign to be attached to the minor which multiplies any constituent in the expanded form, we have only to consider how many displacements would

bring that constituent to the leading place. For example, suppose the determinant (a\b^d^e6) is expanded in terms of its
fourth column, and that
attached to d3 A^ 3
.

it is required to find what sign is to be Here two displacements upwards, and after-

wards three
t

to the left, will bring

d3

he sign

is

negative. This rule

may

to the leading place ; hence be stated simply as follows

:

Proceed from
row,

a\ to the constituent

under consideration along the top

and down

the

column containing the constituent; the number

of

letters
to

passed over before reaching the constituent will decide the

be attached to the minor.

ftiyn

beginning being odd, the required sign is negative. It will be found convenient to retain both notations here em;

at a^

we count a^b^c^d^d^i.Q.

In the example just given five and this number

;

ployed for the development of a determinant. The expansion in terms of the minors, with signs alternately positive and' negative,
is

useful in calculating the value of a determinant by successive reductions to determinants of lower degree. For some purposes,

as will appear in the Articles which follow, it is more convenient to employ the notation first given, in which the signs are all

positive

and the

coefficient of

(whatever the row or column under consideration), any constituent represented by the cor-

responding capital

letter. By substituting for the capital letter the corresponding minor with the proper sign, determined in the manner above explained, the latter notation is changed into the former.

242

Determinants.

1.

a\

b\

c\

Examples.
6.

243
d
e

Expand
c

a
e

/
f
+
b*e~

Ans. (Pd*
1.

c2

/2 -

2 beef- Icafd

labde.

Prove

244
115. Laplace's

Determinants.

Development of a Determinant.
in the preceding Article
of development given
is

The expansion explained in a more general mode

included

by Laplace.

In place of expanding the determinant as a linear function of the constituents of any line, we expand it as a linear function of the minors comprised in any number of lines. Consider, for example, the first two columns (a, b) of

any

determinant

and

let all possible

determinants of the second

order (ap b q ), obtained by taking any two rows of these two columns, be formed. Let the minor formed by suppressing
the ap and b q lines be represented by minant can be expanded in the form

Ap S

then the deter(ap b q )&p , q ,

where

the product of two complementary determinants To prove this, we observe that every term of Art. 113). (see must contain one constituent from the column the determinant

each term

is

a and one from the column

factor ap b q , there

term containing

b. Suppose a term to contain the must then (interchanging p and q) be another the factor - a b hence, the determinant can
q

be expanded in the form 2 (ap b q Ap q and pt q is plainly all of the terms which can be obtained the sum by permuting in every possible way the n - 2 suffixes of the letters c, d, e, &c., viz., Ap,j, the sign being determined in any particular
)
,

instance

by the rule of Art. 108. This reasoning can easily be extended to the case where any number p of columns are taken, and all possible minors formed by taking p rows of these

Each minor is then multiplied by the complementary minor, and the determinant expressed as the sum of all such products with their proper signs.
columns.
EXAMPLES.
Expand the determinant (aibzcsdt) in terms of the minors of the second order formed from the first two columns.
1.

Employing the bracket

notation,

we can write down

the result as follows

where the sign

to be attached to
first

any product

is

involved in the

factor into the positions of first

determined by moving the two rows and second row. Thus, for

example, since three displacements are required to move the second and fourth

rows into these

positions, the sign of the product (a-zb)(c\di) is negative.

Development of Determinants.
2.

245

Expand

similarly the determina

(aibz)

Am.

(03^5)

(ifa)

(c-tdies)

(ffifo)

(2^3*5)

(<iifa) (^2^3*4)

3.

Prove the identity

/2
b\

01

0i

71

/3
02
01

02
03

72
73

Bi
3

71
03

a2

72

00
the

03

03

73

This appears by expanding the determinant in terms of the minors formed from first three columns, for it is evident that all these minors vanish (having one row

at least of ciphers) except one, viz. (flip's).

In general it appears in the same way that if a determinant of the 1m th order 2 contains in any position a square of ciphers, it can be expressed as the product of two determinants of the th order.

4.

Expand the determinant

a

h
b

A
fji

h 9

f
c

/
v

f
u.

in powers of o,

/8,

7,

where

= i'\'

v'\,

7 = Ap -

A' jj..

Ans. aa? + bfp

+ c'f + 2/07 +

2*770

+ 2Ao0.

5. Verify the development of the present Article by showing that it gives in the general case the proper number of terms. Consider the first r columns of a determinant of the n th order. The number of

minors formed from these

r together.
minor), and
will be

is

equal to the
1

number
.

This number multiplied by

1
.

of combinations of n things taken r (the number of terms in each

. .

n
2
.

r (the
3
.
.
.

number
,

of terms in each complementary minor),

found

to give 1

viz. the

number

of terms in the determinant.

116. Development of a Determinant in Products of the leading Constituents. In the present and next fol-

246

Determinants.

lowing Articles will be explained two additional modes of development which will be found useful in the expansion of certain
determinants of special form.

The

application which follows to

the determinant of the fourth order will be sufficient to explain how any determinant may be expanded in products of the lead-

ing constituents
It is required to

expand the determinant

bi

c\

d\

03

D
according to the products of A, B, C, D. In order to give prominence to the leading constituents we have here replaced a\, bz, 03, d by A, B, C, D. When the ext

pansion

is effected it is

plain that the result

must be of the form

-f

Ao +

*2.\'AB

ABCD,
;

where AO
the
the

consists of all the terms in

all

which no leading constituent occurs

;

~2,\A i&

sum sum

of

of all

the terms in which one only of these constituents occurs 'SX'AB is in which the product of a pair of the leading constituents is found
;

It will be the leading term, is the product of all these constituents. observed that the expansion here written contains no terms of the form \"AC, and

and

ABCD,

it is

products of

evident in general that the expanded determinant can contain no terms in which all the leading constituents but one occur, since the coefficient of any
is

such product is the remaining diagonal constituent. It only remains to see what . A', the form of AO, and of the undetermined coefficients A, /u, &# fj.',
.

Putting A, By

C,

D all equal to zero in the identity above written,

we have

dz

Ao

Again, to obtain A,
plainly the determinant

let J5, C,

D be made
cz

equal to zero.

The

coefficient of

is

dz

bz

dz

the coefficient of

is

similarly obtained

by replacing

C,

D each

by zero

in the

Development of Determinants.
iniuor

247
;

complementary to

the coefficient of

let C and D be made zero B and so on. To obtain AB in the resulting determinant is plainly the second minor
;

A.',

The
tlie

coefficient of

any other product

is

obtained in a similar manner.

Finally,

expansion of
bi

A may
ci

be written in the form

d\
dz

as

cz

d
bi
ci

(to

c%

^3

^3

+AC

+ BC

+CV

A determinant
The
panded

whose leading constituents

all

vanish has been called zero-axial.

:

result just obtained

may

be stated as follows

Any

determinant

may

be ex-

in products of the leading constituents, the coefficient of every product in the

result being a zero-axial determinant.

117. Expansion of a Determinant by Products in Pairs of the Constituents of a Row and Column. In what follows we take the first row and first column as those in

terms of which the expansion

ficient, since

is required. This is plainly sufother row and column any may be brought by It will be found convenient displacements into these positions. to write the determinant under consideration in the form

CL\

b\

c\

248
Let
this
.)

Determinants.

be denoted by A', and its leading first minor by the usual notation A. The determinant A' may be said to be derived from A by bordering it, horizontally with the constituents a a, j3, y, ., and vertically with the consti(aib z c3
.

tuents

0>

'j

/3',

7',

...

When

A'
.

is

which contain a are included in aQ A

the
first

expanded, all the terms In addition to this, the ex-

pansion will consist of the product of every other constituent of

column by every other constituent of the first row, such every product of two being multiplied by its proper factor.

What

this factor is in the case of

coefficients of a lf b ly c l5 ...

any product
2,
,

is

easily seen.

^2, expanaccording to the notation explained in Art. 114. It is plain that the factor which multiplies any product, for example aa', in the expansion of A', is the

Let the
sion of

... &o., in the

be Ai, BI,

2y .Z?2,

same

which multiplies a ai with sign changed, viz., the factor which multiplies a )3 is the factor with AI similarly To obtain the factor sign changed of a bi, viz., Bi\ and so on.
as the factor
;

of

any such product the

tioo
is

rule plainly

is

Find
the

the fourth consti-

tuent completing the rectangle

the
constituents

formed by

leading term a
the

and

which enter into the product:

required
so

factor

obtained by substituting for the constituent of

found

the corresponding capital letter ivith the negative sign.

It appears

therefore finally that the expansion of A'

may
-

be written in the

following form

ya

-&c.
Examples
of the utility of this

mode

of expansion will be

found under a subsequent Article.

118.

Addition of Determinants.
any
line

PROP. V.

// every
others,

constituent in

can be resolved into the

sum of two

the determinant can be resolved into the

sum of two

others.

Addition of Deter mrnan is.

249

Suppose the constituents of the first column to be #1 + QI, + a 2 (i A + 3> &c. Substituting these in the expansion of Art. 114, we have
a-,
,

A=

(#1

+ aj Ai +

(a-t

az)

At + (3 +
. .

a3)

+ &c.

+ cizAz + (hA 3 +
or,
01

&c. + diAi + azAz + d 3

+ &c.

+ +

Oi

<7 2

az

which proves the proposition. If a second column consists of the sum of two
easily seen,
first

others, it is

by resolving with reference to one column, and afterwards with reference to the other, that the determinant can
be resolved into the sum of four others.
terminant

For example, the de-

az

is

(using the notation of Art. 113) equal to the

sum

of the four

determinants

Similarly

it

follows that

if

each of the constituents of one

column

consists of the algebraical

sum

of

any number

of terms, of

the determinant can be resolved into a corresponding determinants. For example

a,,

number

a 2 + az
3

Cz

a3
b3

a'3

b3

C3

250
And,
in general, if one

Determinants.

column

consists of the algebraic

sum

of

m others,
sum sum
of

a second column of the

sum

of n others, a third of the

of p others, &c., the determinant can be resolved into the

mnp

&c., others.

Similar results plainly hold with regard to the rows, which may be substituted for columns in the proof just given. 119. PROP. VI. // the constituents of one line are equal to
the

sums of the corresponding

constituents of the other lines multi-

plied by constant factors, the determinant vanishes.

For

it

can then be resolved into the

determinants which separately vanish.

sum of a number For example,

bL

of

nb z

+ n

and each

of the latter determinants vanishes (Ait. 110).

120. PROP. VII.

constituent of

determinant is unchanged when to each any row or column are added those of several other

rows or columns, multiplied respectively by constant factors.

.

For when the determinant

is

resolved into the

sum

of others,

as in Art. 118, the determinants in which the added lines occur

all vanish, since each of them must, when the constant factor removed, contain two identical lines. Thus, for example,
0-1

is

+ mbi +

c\

a?,

<?3

for

when the- second determinant is expressed as the sum of three others, the two arising from the added columns vanish
identically (Art. 119). The proposition of the present Article supplies in practice one of the most useful properties in the evaluation of deter-

minants.

Examples.
EXAMPLES.
1.

251

Show

that the following determinant vanishes

7 + a

out a

Adding the constituents of the second column to those of the first, we can take + j8 + 7 as a factor, and two columns then hecome identical.
Find the value of the determinant

2.

252
in Ex. 7, Art. 112
;

Determinants.
but in general
it

can be effected more readily by direct addi-

tions or subtractions, as in the present instance.

5.

7-205 6-2 -2 0-253

5234

Examples.
8.

253
first

Calculate the determinant formed

:

by the
14
8 2 21

twenty-five natural numbers

arranged in a magic square 10

4

18

22
16 15

12

25
19
13

23
17
11
9.

9
3

24

20

Ans.

-4680000.

Evaluate, by

the method of the present Article, the determinant of Ex. 9,

Art. 114.

11
~

010
y1

A=
yx"

y~

x2 -y2

- 2 -y*

of the following ones.

Here, to obtain the second determinant, we subtract the second column from each In the reduced determinant, subtracting the first row from

each of the following,

1

we
22

find

A=-

-2z 2
x2 -y2 -z~

xz -zz

-y2

-2y2

(y

z2

=10.

(x

+ y + z)
(b

(y

x) (z

+ xa2

y} (x

+y-

z}.

Prove the identity

2 c)

a2

2abc(a

cf.

as a factor.

2 Subtracting the last column from each of the others, (a + b + c) may be taken out the determinant and Calling A', remaining subtracting in it the sum of

the

first

two rows from the

b

last,

we have
a2

+ c -a

254
Adding the
a(b
last
c)

Determinants.
column
a*
to each of the others,

we

ohtain

a2

A =

lab

Multiplication of Determinants.
14.

255

Resolve into linear factors the determinant

abed bade
c

d
c

a
b

The

result is as follows

A = -(a + b + c-\-d)(b + c-a-d)(c + a-b-d)

since each of the factors here written
is

(a

+ b-c-d),
;

a factor of the determinant

for example,

a+

dis shown to he a factor hy adding the second column to the first, and 4 By comparing the sign of a it appears that the subtracting the third and fourth. the to must be attached product. negative sign
It

bc

may be

observed that the determinant of Ex. 9

is

a particular case of the dea,

terminant here considered, viz., that obtained by putting comparing the equivalent forms of Ex. 9, Art. 114.

0, as will

appear by

PROP. VIII. 121. Multiplication of Determinants. The product of two determinants of any order is itself a determinant
of the same order.

We shall prove this for two

The student
is

determinants of the third order.

of the proof, that to
it

will observe,

from the nature

equally applicable in general. product of the two determinants (0i&2

We propose
3 ),

show that the

(aij3 2 7 3 ) is

whose constituents are the sums of the products of the constituents in any row of (0i# 2 c3 ) by the corresponding constituents
in

any row

of (ai|3 2 7 3 ). Since each column consists of the

sum

of three terms, this

determinant can be expanded into the sum of 27 others (Art. Now it will be observed that when any one of these 118).
is

written down, a

common

factor can be taken off each

column

and that

several of the partial determinants will,

when

these

factors are removed,

have two (or more) columns identical. The determinants which do not vanish in this way can be easily
Taking, for example, the
first vertical line

selected.

of the

first

256
column
;

Determinants.

this would give a vanishing determinant if we were to take along with it the first line of the second column. "We take then the second line of the second column, and along with

these two

we must

take the third line of the third column to

obtain a determinant which does not vanish.

Retaining

still

the

first line

of the first column,

we may

take the third line

line of the third

of the second

column along with the second

column.

Taking out the common factors of the columns, write down these two determinants as follows
:

we

c\

cz

Taking in turn each of the other

lines of the first

column,

we

obtain four other determinants which do not vanish.

;

Thus
a
fac-

there are in all six terms

tor in each of these.

and

it is

plain that

(cti

b 2 cs ) is

Taking out

this factor, there

remains the

sum

of six terms

and

this is the determinant

(aif3zjz)is

We

have thus proved

that the determinant above written

the product of the two

given determinants.
either of the given determinants the rows may be written in place of columns; hence, the product may be written

In

in several different forms as a determinant

of course, give the

but these

will,

same value when expanded.

122. Multiplication of Determinants continued. Another mode of proof of the proposition of the last Article, expressing as a determinant the product of two given determinants of the same order, may be derived from Laplace's mode
of development already explained (Art. 115). The nature of this proof will be sufficiently understood from

the application which follows to two determinants of the third order

:

Examples*
The product
of the

257
(01)8273) is (see

two determinants

(ai

2 c3 ),

Ex.

3,

Art. 115)

plainly equal to the determinant

01
flo

03

-1

258

Determinants.

EXAMPLES.
1.

Show

that the product of the

two determinants
a'

a
c

+
+

ib

id
ib

ib'

c'

id'

id

c'

id'

a'

ib'

where

=V

1,

may

he written in the form

V _ ic

B-iA
D + iC
B=
ca'

-B-iA
where

bc'-

b'c

+ ad' cd'

a'd,

c'a

bd'

b'd,

C = ab' - a'b +

e'd,

D = aa' + bV +

cc'

dd';

and hence prove Euler's theorem

(a*

+ +
-

V-

c*

+ dz +

z
)

(a'
2

b'

c'

d' z )

s
+

(aa'

W+
c'a

cc'

dd')
2

(be'

- b'c +
a'b

ad'

a'd)*
2

(ca'

+ bd'-

b'd)

+ (ab' -

+ cd'

c'd)

viz., the product of

two sums each offour squares can

be expressed as the

sum offour

squares.
2.

Prove the [following expression for the square of a determinant of the third
:

order

2 (ac
ac'

b~)

ac'

a'c

2bb'
z
)

ac" + a"c
a'c"

2bb"
.

+ a'c-lbV

2(a'c'-b'
a'c"

+ a"c'

-2b'b"
b" 2 )

ac

"

+ a "c - 2bb"

+ a"c' - 2b'b"

2 (a"c"

This appears hy multiplying the two determinants

b
c
c

-2b
2b'

a
a'

V
b"

c'

c"

c"

-2b"

a"

which
3.

differ only

hy the

factor 2.

Prove the identity

2bc

a*

b*

a2

(a

#3

+ c3 -

2ab-c*

Rectangular Arrays.
This

259

may be

readily proved

by multiplying together the two equivalent deter-

minants

a
b

o
c

-b
c

cab
4.

two determinants of different orders may be multiplied together. may be made equal since the order of any determinant can be increased by adding any number of columns and the same number of rows consisting of units in the diagonal, and all the rest zero constituents. For example,
that

Show

For their orders

may

be written

the only effect of the added constituents being to multiply the determinant by unity. More generally, one set of added constitutents (t. e. those either to the right or the
left of

ing ciphers.

the diagonal) might be taken to be any quantities whatever, the remaining Thus (fli 2 ) may be written in either of the forms
1

260

Determinants.

and, performing on these a process similar to that employed in multiplying two determinants, we obtain the determinant

f3i

c2 ji

+ dz Si

az a z

b z [3

The value

of this is easily found to be

/3 2 )

(a l cz )(a l y z )

i.

sum of the products of all possible determinants which can be formed from one array (by taking a number of columns equal to
e. the

the

number of rows) multiplied by

the other array.

the corresponding determinants

formed from

The
to

student will have no difficulty in extending this proof

any two arrays of the kind here treated. (2). When the number of rows exceeds the number of columns
Take, for example, the two arrays,
ai
bi b*
|
3

the resulting determinant vanishes.

ai
(1),

)3i

02

a2 a3

j3 2

(2).
.

az

j3 3

Performing the process of multiplication, we obtain the determinant

It will be observed that this determinant

arise if a

is

the same as would

column of ciphers were added to each of the given the determinants so formed then multiplied. It and arrays,
follows that the determinant vanishes, since
it is

the product of

two
It
is

factors each equal to zero.

It readily appears that a similar proof applies in general.

as to

only necessary to add to each array columns of ciphers, so make the number of columns equal to the number of rows,

and then multiply the two determinants.

Examples.

261

EXAMPLES.
1
.

From

the two arrays

1 1
1

1
(1),

[1

1
(2),

007
prove a
2.

007

+7

a2

jB

+ 72

From

the

two arrays
a
b
c
(1),
c'

-2*
(2),

c'

2b'

a'

prove
4 (ac 3.
b-} (a'c

- *' 2 - (ac' + a'c - 2bb') z = 4 (be' - b'c)

)

(aV

- a'b) -

(ac'

- a'c) z.

By

squaring the array

a
,' b
c
\

<r

prove

4. Verify,

by squaring the array

abed
of
b'
c'

d'

the result of Ex.

5.

1,

Art. 122.

Prove the determinant identity

(03

This can be proved by multiplying the two arrays

01
2

01

262
124. Solution

Determinants.

of a System of Linear Equations.

114 that a determinant

We have seen in Art.

as a linear

may

be expanded

homogeneous function of the constituents in any row or column, the coefficient of any constituent being the corresponding minor with
its

proper sign.
cizAz

We

have, for example,

A
Now,

= a\Ai +

+ a$Az + &c.

the coefficients

A A
ly

z,

tuents of the other columns

with the consti- 1 identical relations, viz., n by

&c., are connected

biAi +

b z Az

b3

+ &G. =

0,
0, &c.;

c^Ai + CzAz + c$A z + &c. =

for

any one of these

is

what the determinant becomes when the

constituents of the corresponding column are substituted for 2 , 03, &c., and must therefore vanish. 0i,

By

the aid of these relations

we can

write

down

the solution

of a system of linear equations. The following application to the case of three variables is sufficient to explain the general
process.

Let the equations be

0i# + biy + CiZ = m\ a
a-iX

b-^y

cz z

m^

Multiply the
third

by A*

t9 and the equation by AI, the second by and add. The coefficients of y and z vanish, in

first

virtue of the relations above proved

-z

and we obtain

(is

= AS) x niiAi +

m A
z

or

bz

cz

where

stituents, 0j,

represents the determinant formed from the nine conbi, Ci 9 &c.

Linear Homogeneous Equations.

Similarly, multiplying

263

by

B B B
i9 2,

3,

we

obtain

where the determinant on the right-hand

side is

what A becomes

when

m m^ m
l9

are substituted for the constituents of the second

column.

Similarly,

we

obtain for z

As
3

These values

may

be written more compactly, as follows

As =

In general, the values of

lows:

#,

z 9 &c.,

may

be written as fol-

(m^zCz...^)

'

'

where, to obtain the value of any variable, the given quantities m h mz, &c., on the right-hand side of the given equations are to be substituted in A for the coefficients of the variable in
question,

and the determinant

so

125.
linear
ratios

Linear Homogeneous Equations.

formed to be divided by A. When n -

homogeneous equations between n variables are given, the of the variables can be determined by bringing any one of
to the right-hand side of the equations,
;

them

and solving

as in

the previous Article

or

we may determine these ratios more con-

veniently, as follows.

We

equations between four variables, which

illustrate the general process
:

take the particular case of three will be sufficient to

a\x +
<i t

+
bz y
//;//

+ d^r =
cz z
:

x +
.'

+ d2 ic =
+
</.

(1)

+ c ,z

>r

264
To
these

Determinants.

may

be added a fourth equation whose coefficients

are undetermined, viz.,

ax

hy

c4 z

dw =

X.

(2)

Calling (aib z c3 di) as usual A, and solving from these four equations by the method of the last Article, we obtain, since

mi =

0,

0,

0,

X,

the following values

A# =
or,

X^4 4 ,

Ay =

X2?4

As = X<74 ,

Aw

= XJ>4

JL = L. A. B,

JL
Ci

Ji = A' D,

(Ti

The first three of these equations express the ratios of the four variables in terms of the coefficients in the three given
equations.

And,

in general, the variables are proportional to the

coefficients in the expansion of

&

th row of the constituents of the n

supposed added

to the

\.

rows resulting from the given equations.

We can now express the condition that n linear homogeneous

equations should be consistent with one another ; for example, that the equation (2) should, when X = 0, be consistent with the
equations (1). rived from (1),

We have only

to substitute in (2) the ratios de-

when we obtain
#4^4 + b^Bi + CiCt +
c?4

A=

0,

or

A=
The same thing appears from
and
if

0.

the variables do not all vanish,

the equations (3) for A must vanish.

,

if

X =

0,

What

has been proved

is

may

be expressed as follows

The

result of eliminating the variables between

equations in n variables

homogeneous the vanishing of the determinant formed

linear

by the coefficients of the given equations.

quantities &c. (Art. 114), which occur in the expansion of a determinant (i. e. the first minors with their proper signs), may be called inverse constituents; and the determinant

126.

Reciprocal
.

Determinants*

The

Ai

JBi,

Ci

2,

B-i,

formed with them the inverse or

reciprocal determinant.

We

Reciprocal Determinants.

265

proceed to prove certain useful relations connecting the two determinants.

in terms of the given determinant. (1). To express the reciprocal Let the reciprocal of A be denoted by A', and multiply the two

determinants
al
bi

ct

266
whence

Determinants.

B,

ft ft

D,

,A

3 ,

B,
or

thus expressing the

first

minor of A' complementary

minors of
A',

to AI.

Again,

to express the second

we

have,

by an

exactly similar process,

#3

b3

whence

or

Symmetrical Determinants.
proportional
to those

267

of any other row, and the constituents of any col tunu proportional of any other column.
to those

Symmetrical Determinants. Two constituents of determinant are said to be conjugate when one occupies with
127.

reference to the leading constituent the rows as the other does in the columns.
li ;ire

same position in the For example, d2 and

row, and the other the fourth place

conjugates, one occupying the fourth place in the second in the second column. Each

of the leading constituents is its own conjugate. Any two conjugate constituents are situated in a line perpendicular to the

principal diagonal, and at equal distances from

sides.

it

on opposite

one in which every two conjugate constituents are equal to each other. For examples of such determinants the student may refer to Art. 1 14, Exs. 2, 9,
symmetrical determinant
is

10,

and Art. 115, Ex.

4.

In a symmetrical determinant the first minors complementary to any two conjugate constituents are equal, since they differ only by an interchange of rows and columns. The correspondto the

ing inverse constituents are also equal, the signs to be attached minors being the same in both cases. It follows that the
reciprocal of a symmetrical determinant

The leading minors

nants.
*.

symmetrical. are plainly all symmetrical determi-

is itself

The mode

of expansion of Art.

117

is

especially useful in

the case of symmetrical determinants, as will appear from the

examples which follow.

EXAMPLES.
1.

Form

the reciprocal of the symmetrical determinant

i

h
b

A=

f
c

f
Using the capital
Art. 114, so that
letters to

denote the reciprocal constituents, as explained in

A may

be expanded in any one of the forms a

A + hH + ffG,

268 hH +
bB +/F,
:

Determinants.
gG +/F +
cC,

we may

write the reciprocal determinant A'

follows

A
A'

E
B

t>c-f
fg

fg-ch

hf-bg
gh - af

H
G

ch
bg

hf

gh

af

ab

A2

2.

Form

similarly the reciprocal of

h
b

A=

Using a notation similar to that of the preceding example, so that panded indifferently in any of the forms

A may be

ex-

aA + hH + gG +
the reciprocal determinant A'
is

IL,

hH +
by

bB

+/F + mM,
A

&c.,

obtained

replacing in

the constituents by the

if

corresponding capital letters. The student will find no difficulty in writing out, for example, necessary, the expanded form of any of the reciprocal constituents
;

is

the minor complementary

and

F is

proper sign (the negative sign in this case), therefore obtained from the expansion of
its

to/ with

3.

Bringing the

Expand the determinant A of Ex. 10, Art. 114, by the method of Art. 117. last row and last column into the position of first row and first

column, and using the notation of Ex. 1 for the inverse constituents of the leading minor, the result can be written down at once in the form

-A=

A
when both rows and columns
are written in re-

Since a determinant is unaltered

verse order,
if

and
first

the expansion of a determinant be required in terms of the last row last column (as in the present example), it is not necessary to move them in the
instance into the position of
first

row and
it

first

column.

The expansion can be

and
its

written

down from

the determinant as
its

stands, replacing in the rule of Art. 117

last diagonal constituent

the leading constituent and

minor by the

com-

plementary minor.
4.

Expand the determinant

of the above Ex. 2, in terms of the last

row and
using

column, by the method of Art. 117.

Attending to the remark at the end of the preceding example, and

Skew- Symmetric and Skew Determinants.

A, B,
C,

269
3,

F, G,

.ff,

to represent the as follows

:

same quantities as in Exs.

and

the result

may be

written

down

-AP- Bm* -

Cn*

- 2Fmn - IGnl - IHlm.

"When a symmetrical determinant of any order is bordered symmetrically (i. e. by the same constituents horizontally and vertically) the result is plainly a symmetriThe result of Art. 117 shows in general cal determinant of the next higher order.
that the expansion of the bordered determinant consists of the original determinant

by the constituent common to the added row and column, together with a homogeneous function of the second degree of the remaining added constituents.
multiplied
5.

Expand the determinant

A=

This

is

tuent of the added lines being zero.

of the second degree of
a,
,

the determinant of Ex. 2 bordered symmetrically, the common constiThe result is plainly a homogeneous function
7, 8
;

and,

by

aid of the notation of Ex. 2,

may

be

written

down

at once in the

form
C-f

- A = Aa?

+ 2Za5 +
6.

Prove by means of the Proposition of Art. 121, that the square of any deteris

minant

a symmetrical determinant.

128.

Skew-Symmetric and Skew Determinants.

its

A skew-symmetric
is

equal to

one in which every constituent Since each leading con jugate with sign changed.
determinant
is

constituent

is its

own

metric determinant
zero.

all

conjugate, it follows that in a skew-symthe leading diagonal constituents are

A determinant
determinant.

in which all except the leading constituents

is

are equal to their conjugates with sign changed

called a skeiv
is

Thus, while a skew-symmetric determinant

270
zero-axial, a

Determinants.

skew determinant is not. The calculation of a skew determinant may plainly be reduced to that of skewsymmetric determinants by the method of Art. 116.

The remainder

of this Article will be occupied with the proof

of certain useful properties of skew- symmetric determinants. (1). A. skew-symmetric determinant of odd order vanishes.

For any skew-symmetric determinant A is unaltered by changing the columns into rows, and then changing the signs of all the rows. But when the order of the determinant is odd, this process ought to change the sign of A hence A must in this case vanish. For example,
;

b
0.

-a

-b -c
(2).

The

reciprocal of a skeio-symmetric 'determinant of the

is

n th

a symmetric determinant when n determinant when n is even.

order
is

odd,

and a skew-symmetric

In any skew-symmetric determinant the minors corresponding to a pair of conjugate constituents differ by an interchange of rows and columns, and by the signs of all the constituents.

Hence
even.
is

the two minors are equal

is

when

their order is even,

is

namely when n

odd

and equal with opposite signs when n

In the former case, therefore, the reciprocal determinant symmetric and in the latter case it is skew- symmetric, its
;

leading diagonal constituents

all

being skew-symmetric deteris

minants of odd order.

(3).

A skew-symmetric

determinant of even order

a perfect

square.

This follows from the principles established in Art. 126. Take, for example, the determinant of the fourth order,
a
b
c e

-a
-b
-c

-d
-e

-f

Examples.

271
its

and

let

the inverse constituents forming

reciprocal be de-

noted by

A B
19

lf

A*,-&o.

We have then,
A
-/

by

(2),

Art. 126,

AI, and #>, being skew-symmetric determinants of odd and A* = - Bi, since these are conjugate minors ; order, vanish
;

Now

hence y^A = -4 22 , which proves that A is a perfect square. Similarly, for the determinant of the sixth order, it is proved that the product of A by a skew -symmetric determinant of the
fourth order
is

a perfect square

and

since the latter determi-

nant has been just proved to be a perfect square, it follows that A is also. By an exactly similar process, assuming the truth of the Proposition for the determinant of the sixth order, it follows
for one of the eight
;

and

so on.

EXAMPLES.
1. Verify the following expression for the skew -symmetric determinant of the fourth order
:

c
e

-a
-b -d

(af-be

-c
2.

-e -/
skew determinant
x
a
a
b
c

Expand

in powers of x the

d
X*

-b -d
-c -e
When
it is

f
x

-f

the expansion of Art. 116

is

to be observed that the determinants of

employed to calculate a skew determinant, odd order in the expansion all vanish,
as squares.

and those of even order may be expressed

Here the

coefficients of the

odd powers of x plainly vanish, and the result takes the form

A = x*

(a

J2

c2

+ dz +

e*

+/*) x* + (af-

be

272
3.

Determinants.
Expand the skew determinant

A
-a
-b
-c

a,

b
e

d g
i

B
-e

Oh
D

-f -h

-d-g-i-j
The
result

may be

written in the form

z

ABODE + 2rA3C + *Z(ej -fi + ffh) A,

where the
first

includes ten terms similar to the one here written, and the second

the products in pairs of the leading constituents vanish, as also the term not involving these quantities.
includes five terms.
4.

The terms involving

The square

of

any determinant of even order can he expressed

as a

skew-

symmetric determinant.

The The
nants
:

following method of proof is applicable in general. square of (aibzcsdt) is obtained by multiplying the

two following determi-

01

Theorem.
6.

273
A of the
is

Form

the reciprocal of the skew-symmetric determinant

fourth order

in Ex. 1.

Representing by

the function

afbe + cd whose

square

equal to A, and by

A' the required reciprocal,

we

easily find

A'

The value
7.

of this skew-symmetric determinant

1.

may be

written

down by

aid of
.

the result of Ex.

It is thus immediately verified that A'

Form

the reciprocal of the skew- symmetric

= (of- be + cd) z 4 = A 3 determinant A of the fifth order,

obtained

it

by making the leading coefficients all vanish in the determinant of Ex. 3. Since the reciprocal is a symmetric determinant (see (2), Art. 128), and since also must be such that the constituents of any line are proportional to those of any
126),
it

parallel line (Art.

appears that the required determinant must be of the

form
01
2

01 02

0103
0203

0104

0105 0205
0305 0405
05
2

0201
0301 0401 0501
in

02

0204
0304

0302

0402

0403
0503

04

0502

0504

which

0i, 02, 0s, 04,

constituents

whose squares are the values of the

0s are five functions of the second degree in the original five first minors complementary to

the leading constituents of A.

1m +

In general the reciprocal of a skew-symmetric determinant of any odd order 1 is of a form similar to that just written, the diagonal constituents being the

squares, and the remaining constituents the products in pairs, of

1m

-f 1

functions,

each of the

m th

degree in the original constituents.

conclude the subjects of the present the Chapter with following theorem relating to a determinant whose leading first minor vanishes. Adopting the notation of
129.

Theorem.

We

Art. 117,

we regard A as the vanishing determinant, and state the theorem to be proved in the form If a determinant A,
:

whose value

is zero,

be bordered in

determinant so formed by the the product of two linear homogeneous functions of the added constituents.

any manner the product of the leading first minor of A is equal to

',

2 74

De terminan ts.

Eetaining the notation of Art. 117, we propose to prove that the product of A' and Ai may be expressed in the following

form

^A' = -

(A ia +

-Bi/3

+ CVy +

.)(^ia +

A^ + A,y +

.).

(2) of Art. 126 by considering in the determinant reciprocal to A' the values of the constituents

This follows at once from

inverse to a

a',

al

and expressing in terms

of this result

of the original

constituents the determinant of the second order formed

by

these four.

Another proof

from the expansion of Art. 117, by

readily derived the aid of the property of

may be

the reciprocal of a vanishing determinant (Art. 126), viz., that in the determinant formed by constituents lf i9 Ci, &c., the

A S

line are proportional to those in any parallel line. If the original determinant A is symmetrical, and the bordering also symmetrical, the two factors on the right-hand side

in

any

of the above equation

become

identical,

and the theorem takes

the following form: If a symmetrical determinant, whose value is zero, be bordered symmetrically, the product of the determinant
so

leading second minor is equal to the square with a linear homogeneous function of the bordering con" negative sign of

formed by

its

stituents.

Regarding A' as the original determinant, the following useful statement may be given to the theorem just proved If
:

any symmetrical determinant the leading first minor vanish, the determinant itself and its leading second minor have opposite signs.
in

EXAMPLES.
1.

If a skew-symmetric determinant

of odd order

2m +

be bordered in any

manner, the resulting determinant A' is equal to the product of two rational functions each containing the added constituents in the first degree, and the original
constituents in the

th

degree.
7,

Writing, according to the result of Ex. skew-symmetric determinant in the form

01
2

Art. 128, the reciprocal of the given

0102
02
2

0103
0203
03
2

0201
0301

0302

Examples.
and applying the theorem of the present
0i
2

275
we
2

Article,

find

A'=A'

2 (0i a

+ 01030 + 01037+
02)8

)( <J> i

a'+

03017

or
It

=-

(0io

+ 037+
if

.)(i'+

02)8'+
&c., be
(3)

may be

observed that

in this result

a', |8', 7',

made equal

to

- o, -

- y,

&c., respectively,

we

fall

back on the theorem

of Art. 128.

2.

If a skew-symmetric determinant of even order

2m

ner, the resulting determinant is equal to the product of

two

be bordered in any manrational functions, one

of the

m th

and the other of the

(tn

1)**

degree in the constituents.

be derived immediately from the result of the last example by making equal to zero all the added constituents a, ft', 7', &c., except the last, which is to be made= 1. The determinant then reduces to one of the (2m + 1)' ; order of the This

may

kind here considered, the bordering constituents forming the top row and the last column. It appears also that the factor of the m th degree in the result is the square
root of the given skew-symmetric determinant of order 2m.
3.

Resolve into

its

factors

a
a'

7
c

-b
a

ff

-c
b

-a
Ans.

- (aa +

b&

cy)

(ad

cy)

4.

Eesolve into

its

factors

276

Determinants.

MISCELLANEOUS EXAMPLES.
1.

Prove

l\

#2
3

#3
#4

J,

&2

where /has the same

2.

signification as in Art. 37.

Prove

0+7
0'

7+0
7'

o+0
a
"

7
0' 7'

+7'

o'

0'

=2

a
"

0"
3.

+ 7"

7" +

"

"

0"

7"

Prove

07

Py'+P'y

&y'

00

00'

a'0

a'0'

where the

factors

on the right-hand

side are determinants of the second order.

;

Dividing the rows

by 0y,

7'o', a'0'

and putting \

= -,/* =

determinant (omitting a factor) reduces to the form

1
/j.

v
A.

pit

-A
-fi

pv
rA
/*),

v\
A-/*

&c.

A+

/*

-v

Aw

4.

Prove

abed
a'
b'
c'

d'

*.*'. 7 7 7 bed
'd

(b</} (ad'} (ca') (bd'}

(aV) (cd')

abcda'b' c' d'

a' z
a,

b"2

c'

d' z

Multiplying the columns by

treated in
5.

the determinant reduces to the form

Ex.

10, Art. 112.

Prove
2 2 7 + 2

52

07 + a5

(0-7)(a-8H7-a)03-S)(a-0)(7-5).
2

+ 72 S2

o0 + 78

Add

the last column multiplied

by

0075

to the

first.

The determinant

be-

comes then of the form of Ex.

9, Art. 112.

Miscellaneous Examples.
6.

277

Prove

(7
(a
7.

+
-I-

ft

4 5)
4

(7
(o

+ a

J8

5)
2

- 7-

5)

)8

-7-

5)

Prove
a
b b
e

ax + b
bx

ax

bx+c
first

Subtract from the third row the second row plus the
8.

multiplied

by x

Prove similarly
a

278
calculation of a determinant

Determinants.
may
often be simplified

by bordering

it

in this

way.

The determinant

last written is plainly equal to

a\

+
czx

dz
di

Subtracting from the second column the first multiplied by x subtracting then from the third the new second column multiplied by x and, finally, from the fourth the new third column multiplied by x, we have the result above stated.
; ;

10.

Show

that the determinant

\x~

+ cyz + bz*(\-c)xy
(A.

(\2

c)

xy
z

(\
1

- b) xz

Ay + az* + cx (x
1

- b) xz

- a) yz

\zz

bx z

+ cy z -l
is

contains

A.

(x

y*

z2 )

as a factor,

and that the remaining factor

inde-

pendent of A.
\#, \y, \z

1,

Border the determinant, as in Ex. 9, with a first column whose constituents are and with a first row whose constituents are 1, 0, 0, 0. Subtract
;

then x times the

third,

first

and

times the

column from the second, y times the first column from the first column from the fourth. In the determinant thus
first

altered subtract

from the
;

row x times the

second, plus

y times the

third, plus z

times the fourth

1 1.

and the result follows.

Expand

in powers of

x the determinant
d\
z

+x
3

+x

Am.-

12.

Prove the
1

identities
a'

aa'

yy'
55'

155'

Miscellaneous Examples.

279
first

Expanding the first determinant in terms of the minor a formed from the two columns (see Art. 115), we easily prove that it is equal to

A (ft'y
and employing the
Ex.
13.

o'S')

+ S(y a'+

ft' 8')

+ C (a! ft' +

7'

5')

identical equation

A+B+
is

C=

0,

along with the relations of

18, Art. 27, the result follows.

Prove that the determinant of Ex. 12

equal to

7 + o5 70 +
o)3

ft'y'

+
+

o'S'
'5'

7' a'
a'ft'

+ 78

+ y'S'

This follows at once from the relations of Ex. 18, Art. 27. If a', ft', 7', 5' be m put equal to o", ft y, 5'" in the result, we obtain an identity which includes Ex. o, p. 276, as a particular case.
,

14.

Prove

ffi,

02,

#2, c\ being any quantities. 3> *i This follows by subtracting a times the
s

last

column from the

first, ft

times the

last

from the second, &c.

The student

will
th

have no

difficulty in writing
is

down

the

corresponding determinant of the (n

-f })

order which

equal to the polynomial

f(x) whose roots are 01, 02, 03, ... o,,. 15. Resolve into factors the determinant

(0

a')

Here

A=

280

Determinants.

Multiplying the two rectangular arrays

-3a'
(1),
1

3a' 2

-3)8'

3' 2
37
'

-'3

(2),

- 87'

A becomes

equal to the

sum

of four terms, from each of

which we can take out

as a

factor the product of the

two determinants

The remaining

factor is

a#7

2)37 2o'

which can be written

also in the

foim

17.
1

Prove the expansion

Miscellaneous Examples.
19.

281

Each

of the coefficients of

roots as the quotient of

any equation can be expressed in terms of the two determinants.

The student can

easily extend to

any degree the following application to the

equation of the third degree.

From Ex.
xs
a3

10, Art. 112,

we have
I

x*

x
a

a2

Expanding the determinant,

a2 a
1

this identity

can he written

282
21.

Determinants.
Find the condition that the homogeneous quadratic function of three
ax*

variables

by

cz 2

2fyz + Igzx

+ Ihxy

should he resolvable into two factors.

Equating the given function to the product of the factors

(ax

+ 0y +

72)

(ax + ffy +

y'z],

we

readily find

a'
1

Miscellaneous Examples.
By
all

283
;

the preceding example

A must

contain (x

a)

n~ l as a factor
1)

and by adding

the columns
differ

we

see that, it

must also contain x

+ (n

a as a factor.
;

Hence A

can

by a numerical factor only from the product of these the product with the leading term we find

and by comparing

A = (x-a)-

{x

+ (n-

I)

a}.

This result can readily be proved directly without the aid of Ex. 23. 25. The determinant

/i (0)

ft (ft)

/i (7)

/(?)

inwhich/i,/2, /s are any rational integral functions, contains the difference-product

(

7) (7

- o)

(a

as a factor.

This appears readily by reasoning similar to that of Ex. 23.

this nature, in

Determinants of

which the constituents of any column (or row) are functions of the same form, and the constituents of any row (or column) involve the same variable, are called alternants. It is plain that the result is general, and that the alternant
of any order contains as a factor the difference-product of all the variables involved. The determinants of Exs. 9, 10, Art. 112, and Exs. 11, 12, Art. 120, are alternants
of the simplest form.
26. Express in the

form of a determinant the quotient of the alternant in the

preceding example by the difference -product. Assuming, to fix the ideas, that the functions involved are each of the fifth

may

degree (which will include lower degrees by making some coefficients vanish), write a =flio 5 i

we

/2 (a) / (a) =
3

a5

-f

*sa

<?

3 3a

dza?

e^a

Now

taking

a,

to be the roots of the equation

x 3 + pxz

qx

0,
:

and forming the product of the following determinants

a5

284

Determinants.
:

27. Resolve the following determinant into linear factors

01

03
0i
05

03

04
03

05

05
04 03

02
01

04
03

02
01

04 03

05 04

02
01

02

05

Here in all the rows the constituents are the same five quantities taken in circular order, a different one standing first in each row. determinant of this kind is called a circulant. It is convenient to write a circulant in the form here given,

viz.,

such that the same element occupies the diagonal place throughout.

and adding to the sum of the constituents of the remaining columns multiplied by

6 to be

any

root of the equation

x5

0,

first
2

Taking column the

3 4
,
,

0,

re-

spectively,

we

observe that the following are factors of the determinant

,-

0; 01 01 01

+ +

+03

+04
03

+ 05,
04

002
2

05,

# 2 +0*03+ 004
0flr 3

+0 3 05,

+ 0302 +
+
4

+
+
3
,

4+0 2
04

5,

01

02

03

+ 005,
and comparing the
is

the five roots of x 6

5
i

=
it

being

4
;

1, 0,

coefficient of
(cf.

in both expressions

appears that the numerical factor

unity

Ex. 13,

Art. 120).

The method here employed can

n th order
as a product of
0.

a circulant of the easily be extended to express

xn -

n factors by means of the roots of the binomial equation n th order

28. Calculate the determinant of the

Miscellaneous Examples.

285

By aid of this equation the calculation of any determinant is reduced to that of and the values the two next inferior to it in the series A n A n -i, A,,_2, . . A2, Ai
, .
;

of Ai and As are plainly a\ and

a->a\

#2,

respectively.

Dividing the equation just given by

A,,

A n -i we have
+bn
;

An-l

An-l
A,-2

by a similar value the quotient of A,,_i by A w_2, and continuing the proappears that the quotient of any determinant by the one next below it in the series can be expressed as a continued fraction in terms of the given constiOn account of this property determinants of the form here treated are tuents.
replacing
cess, it

called continuants.

"WTien each of the constituents bn , b n -i,

is

63, fa (in

the line

above the diagonal)

equal to +

the resulting determinant

is

a simple continuant.

29. Calculate the determinant of the n th order

a
/3

whose only constituents which do not vanish are a, , 1, lying in the diagonal and the lines adjacent and parallel to it as here represented. The calculation is readily effected for any particular value of n, in a manner
similar to that of the last example,

by

aid of the equation

2 the values of Ai and A2 being a and a - 0, respectively. of the successive values of A, the student will By examining the formation

readily observe that the terms contained in the result are

when n

is

even and of the form 2r

and

when n
will be

is

odd and of the form 2r

1.

For the purposes of a subsequent

investigation, in

which the

results just stated

made use

of, it is

not necessary to

know the
;

coefficients

which enter

into these expressions

general forms of the numerical but such forms can be arrived at

without

difficulty,

and the following general expression obtained for

AM

286
30.

Determinants.
'When a polynomial

U is

divided

by another U'

of lower dimensions, the

coefficients of the quotient,

in terms of the coefficients

and of the remainder, can be expressed as determinants of U and V.

is
'

The method employed

general.

in the following particular case

be of the 5th, and U' of the 3rd degree mainder can then be represented as follows

Let

Z7

equally applicable in the quotient and re-

Q
Also, let

U
From
the identity
:

r
?,'

x+

as

U = Q U' +

we have

the following equations

05

Solving by Art. 124

first

three of these equations

succession,

we determine
-

TO,

q z are expressed as determinants by means of the and taking the first three with each of the others in For example, to find TQ we have from the first r\, r%.
q\,
;

four equations
(to

a\

tto'

a\

=
a\
^2'
(to'

0,

or

#2

ft

#3

+ ?"o
coefficients of the quotient,

^3

31.

Find the general forms of the

and of the re-

mainder,

when a polynomial
a;

of even degree

2m

is

divided by a quadratic.

Taking

a#

as the given quadratic function,

we have

the identity

Writing down the

solve for go, qi, qi,
.
.

first
.

qr,

it is

equations, formed as in the preceding example, to easily seen that the value of q r thence derived is

with the constituents

a determinant of the r th order of the form treated in Ex. 29, bordered at the top ar . 1, 0, ... 0, o, and at the right-hand side with OQ, i,
. . .

Expanding

this determinant in
is

terms of the

last

column,

any quotient the form

expressed
qr - ar

by means

of a series

immediately seen that of the determinants of Ex. 29 in

it is

a,-i

A\ + a r-z Aa - &c.

A r_i

A,-,

Miscellaneous Examples.
the upper or lower sign to be used according as r is even or odd. coefficients of the remainder, \ve have the equations

287
To
obtain the

Expressing the values of #2m-3,

2w-2

by the formula

just proved,

to the results of Ex. 29, \ve derive the following general forms for TQ

and attending and r\

:

+x

288
Any equation
real.

Determinants.
in the series, being of the

same form

as

A =
;

0,

has

all its

roots

It is plain also that each of these equations is a limiting equation (see Art.
it

83) with reference to the equation next above

in the series

since, in order that

a change of sign

may

be

lost

between

An

and

A M-i

at the passage

through each of

two consecutive roots these two values of x.

has been just proved
the equation
A,,-i

of the former, the value of

A n -i

must, change sign between

it

The equation A n = appears that when

-- 1

may have

this equation

equal roots, and by what has r roots equal to a,

has r

roots equal to o, the equation

A n -z =

has r

roots equal to a,

and so on.

matics and physics.

The determinant here discussed occurs in several investigations in pure matheThe proof here given of the property above stated is taken from Salmon's Higher Algebra (Art. 46), to which work the student is referred for other
33. If the determinant of the preceding

proofs of the same theorem.

that every

first

minor has r

roots equal to a

example have r roots equal to o prove 2 roots every second minor r

;

equal to

a,

and so on.
. . .

Employing the notation A, H, G,

minant,

for the elements of the reciprocal deter-

we have

the equation

AB - H* =
Now
leading
it is

An-z

Aw

easily seen

by proper

transpositions of rows and columns that every

minor contains the multiple root r 1 times. It follows from the must contain this root r 1 times and equation just written that the minor may be taken to represent any first minor.
first

34.

Find the conditions that the equation

a

+x
h
b

+x

f
c

+x we
readily derive the

should have equal roots. Since each first minor must contain the double root,
required conditions in the following form
:

This and the preceding example are taken from Eouth's Dynamics of a System
of Rigid Bodies, Part n., Art. 61.

CHAPTER

XII.

SYMMETRIC FUNCTIONS OF THE ROOTS.

130. Hfewton's

Theorem on
now resume

the

Sums of the Powers

of the Roots.

We

the discussion of symmetric

functions of the roots of an equation, of which a short account has been previously given (see Art. 27) and proceed to prove
;

certain general propositions relating to these functions : PROP. I. The sums of the similar powers of the roots of an
equation can be expressed rationally in terms of the coefficients.

Let the equation be = X + piX*- + pzX*-2 + f(x] = (x- cti) (x - a 2 ) (x - a 3 )

11

+ pn
(x
. .

a) =

0.

; or, adopting proceed to calculate Sa , Sa , sm in terms of the coefficients the usual notation, s2 , 3 ,
.

We

Sa w

p p^
l9

...

pn

We have,
x
= nx

by Art.
x
(n

72,

a\

a?

n~ l

n~z

l)pix

(n

of Art. 8,

and we
x
a

find, dividing

by the method
+
a*

= xn ~ + a
l

+a

290

Symmetric Functions of

the Roots.

*!,

If in this equation we replace a by each of the quantities + a/, o 2 ... a n in succession, and put sp = Sap = a-P + a zp +
,
. . .

we

have,

by adding

all these results,

the following value for

f(x) = nx"- +
1

5:

..+

s n_ L

whence, comparing this value of /'(#) with the former,

obtain the following relations:
i

we

+Pi =
+
piSi

0,

s2

+ 2p 2 =

0,

0,

0,

0.

The

first
s2
;

the second

pn ; equation determines Si in terms of p l9 p^ the third s3 and so on, until s_i is determined.
.

We find in this way

= - Pi 5

2 - 5 (p z p^p + 5 (p z p3 - p5)
l

&c.

Having shown how

terms of the
coefficients,

Si,

sz , s3 ,

s n_ l

can be calculated in
to extend our results

we proceed now

Neivtorfs Theorem on
to the

Sums of Powers of Roots.

.

291
..

sums

of all positive powers of the roots, viz., s n , s n+1 ,

sm .

For

this

purpose

we have
^ x + Pizm ~ l + Ptxm ~ z +
. . .

a*/
sion,

(a?)

pn xm~*.
. .

Eeplacing, in this identity, x by the roots a

t,

a 2,

a n , in succes-

and adding, we have

m +PlS m-i

+PiS^
we
2

+ p n S,n- n =
1,

0.

Now, giving

w
+

the values n, n +
s

n +

2, &c., successively,

and observing that

sn

obtain from the last equation

.

/?i*-i

+^

s^2 + + +

+ np n

0,
0, 0,

+ jn*i =
+ ^n*z =

&C.

Hence the sums

expressed by transforming the equation into one whose roots are the recio, and applying the above formulas, procals of ai, a 2 a 3
, ,
. . .

of all positive powers of the roots may be And by integral functions of the coefficients.

we may
roots of

express similarly all negative powers of the roots.

131. PROP.

II. Every rational symmetric function of the an algebraic equation can be expressed rationally in terms

of the

coefficients.

It is sufficient to prove this

functions, since fractional

theorem for integral symmetric symmetric functions can be reduced to

. .
.

symmetric functions.
is

a single fraction whose numerator and denominator are integral a Every integral function of ai, a 2 ,
the

sum

of a

number

of terms of the
;

form Na^ p a z9 a 3r
.

where

and if this function is symmetrical we can write it under the form N'S.a^ a zq ai all the terms being of the same type. Therefore, if we prove that this quanJV is a numerical constant
. .

tity

can be expressed rationally in terms of the

coefficients, the

theorem will be demonstrated.

We
pq -

shall first establish the

:

p ? following value of the symmetric function 2ai a 2

p+q .

(1)

292
To prove

Symmetric Functions of
this,

the Roots.

we multiply
Sp
sq

together sp and
. . .

s q9

where

= af + a/ + a/ + =
ai
?

a/,
a n?
;

+ a zq +
...

a 3?

whence
SpSq

= aj* 9 + a/+ q +

+ anp+ q + afa

+ ajaf + &C.,

or
SpSq

Sp+q

2WW,
,

which expresses the double function Set/ a zq in terms of the single functions sp9 s q sp+q in the form above written.

We proceed now to
function,
i.e.,

prove a similar expression for the triple

Sp S q Sr

Sq+rSp

Sr+p S q

Sp+q Sr

+ 2sp+q ^.

(2)

Multiplying together

Saip a 29 and

s r9

where

Sr

= a* + a zr + a 3r +

+ a/,
viz.,

we

obtain an expression consisting of three different parts,

terms of the form

Sa^+ W,

Sa^a/, and

Hence

a formula connecting double and triple symmetric functions.

But, by
(1),
p +r?

p+q+r ,

q+r
i

a/ =

S+t-S

Substituting

these

values,

we

find

the triple

function

/ expressed as the series Si, s 2 , s 3 , &c.

above in terms of single functions in

In the same manner the quadruple function Saip a zq a 3r a/

Newton's Theorem on Sums of Powers of Roots.

can be made to depend on the triple function
ultimately on
s lt s2 ,

293
and

2a^

a zq a/,

s3 ,

&o.

and

so

on.

Whence,

finally,

every rational symmetric function of the roots may be expressed in terms of the coefficients, since, by Prop. L, i, s2 $3, &c., can be so expressed.
,

The formulas (1) and (2) require to be modified when any of the exponents become equal. p q Thus, if p = q, ai a z = a/ar , and the terms in (1) become
2

equal two and two

therefore

2aip a 29 =

2Sa/a/

whence

= r in Sa/a 2? a 3r the six terms obtained r by interchanging the roots in af a zq a 3 become all equal hence
Similarly,
if

p =

2s3p ).

And,

in general,
.

if

exponents become equal, each term

is

repeated 1

times.

EXAMPLES.
1.

Prove
2oiPo2 5a 3 r a4 s

sp s q s r s t

2sp s<2 s r+s

2.

Prove
2

m 04 n =

6* m

132. PROP.
Piipi,
Pn>
*s

III.

The value of
r

sr ,

expressed in

terms of

the coefficient of y
v
).

in the

expansion by ascending

powers of y of
Since

n - r logfy /f

xn + p^xn
putting

-1

+ p 2 x n -~ +

+ pn =

(x

oj (x

- a2

(x

- an

),

- for x in

this identical equation,

we

find
... (1

1 + p\y +PZ!/~

3 +p*y +

+ p n yn -

(1

-ai^)(l

~^y\

y).

294

Symmetric Functions of

the Roots.

Now, taking
y*

the Napierian logarithms of both sides,

+p*

Coefficients in

Terms of Sums of Powers of Roots.

295

which becomes by expansion

1-

--s
2
1

**-4.

1.2.3

Now, comparing we
and we
see

the coefficients of the different powers of y,

obtain values for JPI,

p^ p^

pn

in terms of

s lf s 2 ,
.

sn

that/V involves beyond sr If the identity (1) be differentiated with regard to y, the equations of Art. 130 connecting the coefficients and sums of

no sum

of powers

powers

may

be derived immediately from the resulting identity.

It is important to observe that the

problem to express any

is

symmetric

function of the roots in terms of the coefficients or

any

coefficient in

perfectly definite, there

terms of the sums of the powers of the roots being only one solution in each case.

"We add some examples depending on the principles established in the preceding propositions.

EXAMPLES.
1.

Determine the value of

where

01,

o2

03,

an

are the roots off(x)

0,

and

<t>(x) is

any

rational

and

integral function of x.

We have
/'(*)
1 1
1
0.2

01

an

and
f(x)

01

*(*)
|

02

X - On

296

Symmetric Functions of

the Roots.

equation,

Performing the division, and retaining only the remainders on both sides of this we have
(02)
ft

(an)
'

f(x)

x-a\

a?

.OB

z-a n
- 03)
. .
.

whence

x- 1 + Hi xn ~z +.... + -R-i = 2ft (ai)

and, .comparing the coefficients of x
n~ l

(a:

- 02)

(s

(x

- a)

on both

sides of this equation,

/2. Prove that sp

is

the coefficient of

- in the quotient of the division of f(x)

x.

by /(a?) arranged according to negative

3.

powers of

the coefficient (with sign changed) of %P~ I in the same quotient arranged according to positive powers of x. ^ ^ / ^'" 4. If the degree of ft(#) does not exceed n 2, prove ,/T",

Prove that

s.p is

where

2
r=l

denotes the

sum

obtained

by giving

r all values from

to

inclusive.

"We have, by partial fractions,

r
az

f(x)

x -

01

a;

An
x

on
... in succession,

and, multiplying across by /(#), and putting x equal to

01, 02,

f(x)
whence

/'(ai)

x -

01

/'(o2 ) *

a2

/'

(a,,)

a;

'

a,,

"When <(#)
function ction of -,

is of

the degree n

expressing the
is

first side

of the equation as a
as a multiplier

it

readily appears that there

coefficients,

no term without x

"We have therefore, comparing

As may be any rational and integral function of degree we have the following particular cases which are worthy of
</>

not higher than n

special notice
1
:

2,

2-

ow

"2

=0,

on ~ 3

=0,

...2--=0, 2

-=0.

Examples.
Prove that the sum of
the quantities
01, 02,

297
Ur
,

all
,

the homogeneous products

is

of the r th degree of

... a n

equal to

"\Ve

have, putting y

=
1

-,

am

(1-aiyMl-

Also

/(*)/()*-'
and therefore

whence, comparing

coefficients of

r y in these two expansions,

6. To express the coefficients of an equation in terms of the homogeneous pro ducts of the roots, and vice versa
.

From

the equation of the preceding example

we have
which gives the following relations
pi
4:

Hi

0,

0,
0,

&C.
n, p\, pz,
.

These equations

(in

which, for values of r not greater than

pr

and Hi,
Hi,
ri2,
.

IIs,
.
.

nr
and

are interchangeable) determine pi, pz,

vice versa.

Pn in terms

of

Iln,

By means
tric

of this

and the preceding example the values of the following symmein terms of the coefficients
:

functions

may he found

7.

To

express

n r by

the sums of the powers of the roots.

(1

Representing by
ing,

- the product

- ai y) (1 - a2 y)

(1

- ay), and

differentiat-

we

find

\du

= 2
1

udy
also

ay
Iliy
-f

298
We

Symmetric Functions of
have, therefore,

the Roots.

Now
number
HI,

comparing the several

of equations
. . .

coefficients of the different

by means

of

powers of y, we have a which the sums of the homogeneous products

s\, s 2 , $3,

112, IIs,

may be

expressed in terms of

&c.

8.

To

find a general expression for sm in terms of the coefficients p\,

pz

pn

of

an equation of the n th degree. We have

s *v z

3 S o 3*/ 6

+ ...+

(p\y
of

Now, making use of the known form of the coefficient of y m in the expansion of + Pny n r by the multinomial theorem, and comparing coefficients + pzyz + m y in the above equation, we find
)

in

which

ri+
r\

r2
2r-2

+ +

r3

rn

3r5

+ nrn = m

and n, r2

r%,

rn are to be given

all positive

satisfy the last of these

tegers,

two equations.
r(r f +.l)

Also, representing

integer values, zero included, which any of these inby

.2.3 ...n,

with the assumption that r(l)

9.

=
. .

when
sm .

n=

0.

To

find a general expression for

si, s%,
.

any

coefficient

pm

in terms of the

sums of

the powers of the roots

We have

When
example,

the factors on the right-hand side of this equation are developed, and the

coefficients of

y on both

sides compared,

we

find,

employing the notation of the last

r(n +

1)

r(r 2 +

1)

r(r m

1)

2^

3'-3

mW
which

in

which n, r2

... r w are to be given all positive values, zero included,

satisfy the equation

3/ 3
-

+ mr m =

m.

Theorem.
134. Definitions.

299
The
weight of

Theorem.
is

any sym-

metric function of the roots

term in the function.

six.

For

the degree in all the roots of any 2 3 example, the weight of 2a/3 7 is
of the roots

The

order of

any symmetric function

is

the

highest degree in

which each root enters the function.

For

2 3 example, the order of 2aj3 y is three. It has been proved (see Art. 28), that the weight of any

symmetric function of the

efficients
ff
j

rti>

#2}

tf> is

roots, when expressed by the cothe same as the sum of the suffixes

of each

term in the expression.

viz.
:

We

now prove

another im-

portant theorem,

If any symmetric Junction

/>i,

be expressed in terms of the coefficients

is

p:, the

Pn,

the degree in the coefficients

the

same as

the order

of

symmetric function.

For example, Sa 2 /3 2 =p*2 -

%p\p* + 2/ 4 ,
coefficients,

no term being of higher degree than the second in the and the order of the symmetric function being two.

The student may

easily satisfy himself of the truth of this

theorem by observing that in the equations (2) of Art. 23, the value of each coefficient in terms of the roots contains each root
in the
first

efficients will

power only hence the highest degree in the cobe the same as the highest degree of the cor;

responding symmetric function in any individual root. add the following formal proof, as it is in accordance with the
proofs of certain general propositions to be given subsequently.

We

Replace the

coefficients

p p
l9

z,

p n by
#0

--,

#0

#0

Now,

if

(m, o 2 ,

o n ) denote
roots,

any

rational

and integral

symmetric function of the

w
flro

we have
,

0(ai, a 2 ,

an)

= F(a

a l9 a z ,

),

where

the degree in the coefficients of F(a tfi, tf-, ...), a homogeneous and integral function of the coefficients, not
OT is
,

divisible

by

We

require

now

to

show that

vr is

the order of 0.

For

this

300

Symmetric Functions of

the Eoots.

purpose change the roots into their reciprocals, and, therefore, a Whence #oj i,... a n into a nt a n -i,
.
.

^n
also

w
f

,,...
J

= F(a n

_!,

#n_2 ,

(1)

J^
where p
sible
is

\l/

(a

a2 q 3
, .
.

g n)
'

an ) p

the order of 0, and

^ an
;

integral function not divi(a!a 2 a 3

.

by

the product of all the roots

lowest
(1),

common denominator we have

a 2 ...
,

of all the terms.

a n } p being the Substituting in

an)=a nP-F(an
p
a2,
.
.

an. ly

).

From
product
ficients

this equation it follows that

( ai ,
.

is

equal to

for if

were greater than w, $

ai a 2
.

an)

would be

divisible

by the
which

an,
. . .

and
)

if it

were

less,

the function of the coef-

F(a m an_!,

would be

divisible

by a n both
,

of

suppositions are contrary to hypothesis.

135. Calculation of Symmetric Functions of the Roots. Any rational symmetric function can be calculated by the method of Art. 131. In practice, however, other methods are usually more convenient, as will appear from the examples given at the end of the present Article, and from the following
Articles, in

which we shall give certain general propositions

cases facilitate the calculation of

which in
functions.

many

symmetric

The number
is

of terms in

easily determined.

3 2 2ai a 2 <* 3 of the equation 1) degree being the number of permutations of n things taken three If the exponents of the roots in any term be not all together.

any symmetric function of the roots For example, the number of terms in of the n th is n(n (n 2), this

different, the

number

of terms will be reduced.

for a biquadratic consists of twelve terms only (see and not of twenty-four, since the two permutations a/3y, ay/3

2 Thus, 2a j3y Ex. 6, p. 48),

give only one distinct term,

viz.,

2 a /3y, in

Sa^y.

The student

Calculation of Symmetric Functions.

301

culty in

acquainted with the theory of permutations will have no diffieffecting these reductions in any particular case.

When

two exponents of roots are equal, the number obtained on the supposition that they are all unequal is to be divided by 1.2; when three become equal this number is to be divided

by 1.2.3; and so on. In general, the number of terms in Sa ^a 2 9 a 3r ... of the equation of the n th degree, each term containing m roots, and v of the indices being equal, is
1

1.2.3.

.v

When the highest power in which any one root enters into the symmetric function is small, i. e., when the order of the function (see Art. 134) is low, the methods already illustrated
in Art. 27
of the

be employed with advantage for the calculation symmetric function of the roots in terms of the co-

may

efficients.

It

is

important to observe that when any symmetric function

all

whose degree in
n th degree, numerical
its

the roots

(i.
l9

e., its
2
.

in terms of the coefficients

p p

pn

weight) is ??, is calculated for the equation of the

coefficients

value for an equation of any higher degree (the being all equal to unity) is precisely the

plain that no coefficient beyond p n can enter into this value, and the equations of Art. 130, by means of which the calculation can be supposed to be made, have precisely the

same

for

it is

same form

an equation of the n th degree as for equations of It is also evident that the value of the same all higher degrees. for an equation of a degree m (lower than n) symmetric function
for

is

obtained by putting

p m+ p m+Z9
i,

pn

all

equal to zero in the

th calculated value for an equation of the n degree, since the can be lower derived from that of the n th by of degree equation

putting the coefficients beyond p m equal to zero and the corresponding symmetric function reduces similarly by putting the a n each equal to zero. roots a OT .u, a m+2
;
. .

302

Symmetric Functions of

the Roots.

EXAMPLES.
1.

Calculate

2oi 2 0203

of the roots of the equation

l '

+ piX n ~ +p Z ^ n Z + ....+ pn-lX

Multiply together the equations

-f

pn =

0.

In the product the term

2 ai 03 03 occurs

only once

the term 01 02 03 04 occurs four

times, arising from the product of ai

by

020304, of 02

by

010304, of 03

by

010204,

and of

04

by

010303.

Hence
2oi 2 0203

4501020304
4j94.

=
(Compare Ex.
of Art. 131,
2
*4,

therefore

2oi 2 0203

= pips

6,

Art. 27.)

If the calculation were conducted

by the method
3

we

should have

201

03

= ^*2l -

*l3 - i*2 +

which

leads,

evident that in this case the former process

s\, so,

on substituting the values of Art. 130, to the same result but it is is much more simple, since the values of
;

&c., introduce a

number

of terms

which destroy one another.

2.

Calculate

2oi 02 2
2

for the general equation.

Squaring
20102 =$2,

we have
2oi 2 02
2

22oi 2 0203

6501020304 =J?2 2

In squaring
0102

it is

evident that the term 01 02 03 04 will arise from the product of

by

03 04, or of 01 03
is 6,

by 02 04,

or of ai 04

by

02 03

hence the

coefficient of ai 02 03 04

in the result

since each of these occurs twice in the square.

8,

The

result differs

from the similar equation of Ex.

01020304.

Art. 27, only in having

before the term

Hence,

finally,

3.

Calculate

2 or3 02

for the general equation.

9, Art. 27,

We have, as in Ex.

Hence, employing previous

results,

4.

Calculate 2oi 2 02 2 03 for the general equation.

result will be the

The

same as

if

the calculation were

made

for the equation of

the fifth degree.

Examples.
To

30r3

consider

obtain the symmetric function we multiply together 20103 and 2010303 and what types of terms, involving the five roots 01, 02, 03, 04, 05, can result.
;

The term or 02'- 03 will occur only once in the product, since it can only arise by Terms of the type oi 2 02 03 04 will occur, each three multiplying 0102 by 010203. times since or 02 03 04 will arise from the product of 01 02 by 01 03 04, of 01 03 by
;

010:04, or of 0104
ai 02 03 04 05 will

by

010203

and

it

cannot arise in any other way.

The term

occur ten times in the product, since it will arise from the product of any pair by the other three roots, and there are ten combinations in pairs of the five roots. We have then, for the general equation,

2aiO22aiO2 03

2oi O2~O3

32ai 2 a2O3O4

-f

1020102030405.

[We can verify this equation when = 5, just as in Ex. 9, Art. 27 ; for the product of two factors, each consisting of 10 terms, will contain 100 terms. These are made up of the 30 terms contained in 2oi 2 o2 2 03, along with the 20 terms contained in
times.]
2oi'-

020304, each taken three times,

and the term 0102030405 taken 10

Thus the

calculation of the

required symmetric function involves that of

04; for

which we

easily find

201 201020304

201 2 020304

520102030105.

Hence,

finally,

we

obtain

The

process of Art. 131

would involve the calculation of

$5;

and many terms

result.

would be introduced through the values of si, $2, &c., which disappear in the 2 5. Find the value of 2or as 03 04 for the general equation.

We multiply together 2 ai 02 and 201020304, and consider what types of terms can arise involving the six roots 01, o 2 03, 04, 05, oe- The term oi 2 022 03 04 can occur only once. Terms of the type 0^0203 04 05 will each occur four times, this term
,

arising

from the product of 0102 by 01030405, or of 0103 by 01030405, or of 0104 by

01020305, or of 0105

by 01030304.

The term"oi0203a4050j

will occur 15 times, this

being the number of combinations in pairs of the six roots.

201 02 201 OJ 03 04

Hence
2fll 02 03 04 05 06-

2oi 02 03 04

4 2oi 2 02 0^04 05 +

We have again,
Hence,
finally,

for the calculation of 2oi 2 {o2030405,

2oi 2oi 02 03 04 05
201 2 02
6.
2

=
3

2oi 03 03 04 05

62oi 02 03 04 05 O6.

04

= Pip* -

4PIP5

9jJ 6 -

Find the value of 2oi 2 02 2 aj2 in terms of the

coefficients of the general

equa-

tion.

Here, squaring 2010303,

we have
20 2oi 03 03 04 05 ae,

2 2 2 2 2 2oi 02 03 2oi oj 03 = 2oi 02 as + 2 2ar oz 03 04 + 6 2oi 03 03 04 05 4

from which we obtain

2o;
2 2 -03-

= #J2 -

2^j Pi

304

Symmetric Functions of

the Roots.

136. Brioschi's Differential

Equation.

M.

Brioschi

has given the following differential equation connecting the coefficients and sums of powers of the roots of an equation
:

dr+k

To prove

this

we
.

have, as in Art. 132,

.

+ Pnl/ n )^-l/s

r sr --^s --fs ..--y A O T

2 s
.

and

differentiating,

(1
uSf

+P& +p

z y +

+pn y

n
)

s - (1 +p,y +p,y*+.

+pn y

n
)

y
I

whence, comparing the

-4~ =
as r

coefficients of the different

powers of

y,

0,

when

< r

dp r

dp r+k

We can now express the result of differentiating with respect

to
sr

any function

of the coefficients

Since
dpi dp*

all vanish,

d _.
i
-*-

dF dpr
7
7
i

dF
7

dp r+l
7

V /^l)

1J &)

A'oj

J n)

dF

o?sr

c?jt?r

^s/-

dp r+l dsr

dp dp n dsr

and, applying the formula given above, this reduces to

IfdF
,

dF
dp r+l

dF
...
n -r

dF'
,

dp r

dp r+2

dp n

By

means

of this result symmetric functions can often be

calculated with great facility, as will appear from the following

examples

Examples.
EXAMPLES.
1
.

305

Calculate the value of the symmetric function

2oi

2
e

2
2

as en

of the roots of

the equation

SP

+ p\ a^- 1 + pzx n ~~ +

+ pn -

0.

Knowing
order,

the literal part of

the order and weight of any symmetric function, -we can write down Here 2 is of the second its value in terms of the coefficients.

and

its

weight

is

eight
tops
-f

hence
tipipi

2=
where
to, ti,
2,

t2

p 6 p2 + hpsps +

tipi*,

&c., are numerical coefficients to be determined.

,

Terms such as p 6 pi*, pspipz, j0s.Pi 3 &c., although of the right weight, are an order, and therefore cannot enter into the expression for 2. Again, 2 expressed in terms of the sums of the powers of the roots is of the form r expressed in terms of the sums of for, in general, 2ai*> o 2 as -F(s2, 4, *e, SB)
of too high
;

sp sp + q sp ^ qr skp are even therefore in which are sums of even powers when p, q,r, case none hut even sums of powers enter into the expression for 2.
is
, , , . . .

the powers of the roots,

all of

made up

of terms such as
.

this

rf2

Also, since
dss

=0, and

d"S,

= 0, we have,

dF
,

using the formula above given for

asr

ds^

tops
to pi

+
+
tipi

=
we

0.

From

these equations
*o

infer

*i

0,

tz

t3

0,

tz

0,

*i -f

but

4=1,

since for a quartic

ti

2 =p
^o =

therefore
^3

=-

2,

2,

=-

2,

tz

and, substituting these values of

to, t\, tz,

h, t,

2.

Calculate

2ai 2 a 22 o3

for the

same equation.
Zpzpi +
2 J?s
.

Am. - 2pe +
3.

2p\p 5

(Compare Ex.

6,

Art. 135.)

Calculate for the

is

Here the weight

2oi o2 a3
3

same equation the symmetric function 2oi 3 o 2 2 03. hence six, and the order three
;

= top G +

tip 5 pi + tzpipz

t3

pipi

+ <4^32 +

ts

Also 2, expressed in terms of

*i, >z, *3,

&c.,

is (see

Art. 131),

Now,
regard to

differentiating
6,

by means

of Brioschi's equation these

two values

of

2 witli

and comparing

differential coefficients,

we have

306

Symmetric Functions of
$5,

the Roots.

Differentiating with regard to

we have
5s t

topi +

t\p\

=
$4,

=-

5pi

.-.

7.

Differentiating with regard to

top*

+ tip? +
t

tzp-t

tap?
8,

4s 2

(p? -

2pz)

whence

tz

== if

h+
tz

t3

4;

and
*3

3,

4.

; Again, taking the particular case

for

vanishes

2 roots vanish.
;

And we

find

and ^ by

when n

3 roots vanish

for in this case

and therefore
whence,
finally,

U=-

3,

fe

137.

Derivation of

new Symmetric Functions from

relation such as

a given one.
a w

From any
a2
,

S0 (cti,

a n) =

F(a

a i9 a z

a n ),

where ^

is

an integral function, of the order

r,

of

some or

all of

the roots of the equation

a
a

+ na xn
l

~l

n +

y
J.

f/ 2
.

^~ 2 +

+ an =

0,

we may

derive a

number

of other

symmetric functions and their

equivalents in terms of the coefficients. For this purpose diminish each of the roots
x,

by any quantity

and consequently change any

this is

coefficient ar into

Ur (see Art. 35).

Z72 ,
.
.

When

done the original relation becomes

-x, o 2
-ar,
.
. .

a n -x) =

F(U^
a

Z7i,

Un

) ;

and comparing the

coefficients of the different

powers of x on
of symmetric

both sides of this equation,

we have

number

functions of the roots expressed in terms of the coefficients as It should be observed, however, that this method leads required.
to

no new symmetric functions when the given function

is

function of the differences of the roots.

Equation of Operation.

307

138. Equation of Operation. We now proceed to deduce an important equation of operation in the notation of the differential calculus, which may be applied to furnish the
results of the last Article.

Let

tf.rtf)

(ai,

a2,

a) = F(a Q

a lt

a,, ...

a n ),

as in the last Article.

Adopting the notation

_,

aai

< '7-* + ... + ua

aa.2

#
we have

d
flo

-T-

da

+ 20!

d
da z

d rt + 302 -- +
da3

+ na n ^
:

d
da
, tl

the following equation of operation

SafQ

( ai ,

a2,

a) = DF(a

a l9

a n ).

To prove
ff

this,

we

have, as in Art. 137,

. . .

[ff

tf>(ai-#, a 2

-#,

a n -x) =

and,

by Taylor's theorem,

where
Again, omitting

^ = ^!(u
all

"2?

powers of
, i

a?

higher than the

o

first,

F( U
or,

Ui,

...

Un

becomeslF (

when expanded,

F + x(a<
\

+2a
ddi

+
da?

...

+ na n ^

da n

where

F(a

ah

an]

whence, comparing the equation above written,

coefficients of
viz.,
.

x in both expansions, we find

^^(a,,
from

2,

O n)
if

= DF(a^

ft,

. .

tf

This equation shows that

by

the operation

S, its

a symmetric function be derived value in terms of the coefficients

x2

308

Symmetric Functions of

the Roots.

may
tion

be derived from the corresponding value of

by the opera-

D.
<ty

Again, since

this equation, a^tffy

and DF may take the place of and F in becomes D*F, &c. It may be noticed,
, ,

2 3 S moreover, that if S0 vanishes, S &c., thus that x disappears in the expansion of

all

vanish; and

0(oi-#, o 2

a?,

o -

x}.

Now
af*F(a0t

this can

differences
i,

of
2,

01, a 2

is a function of the happen only when .... a n whence we conclude that if

,

...

is

the value in terms of the coefficients

of a function of the differences of the roots, then

DF(a
vanishes identically.

0i,

2,

W)

This identical relation

numerical
coefficients in

is

often sufficient to determine the

by the
is

coefficients,

a function of the differences expressed when the order and weight are known. It

not sufficient for this purpose when there exist more than one function of the differences of the required order and weight.

We
this

add examples way.

of functions of the differences determined in

EXAMPLES.
1.

Determine a function of the differences whose order and weight are both
>

three.

Assume

=A

these being the only three terms which satisfy the required conditions. It is that the operation is performed by applying to the suffix evident from the form of

of any coefficient a r the same process as in ordinary differentiation index. Thus Dar = ra r-i, and therefore

is

applied to the

=
Hence

0.

ZA + E=0,
and putting

and

2B + 3(7=0;

we have

B = -3,
whence,
finally,

and

(7=2;
G.
(See Art. 36.)

Operation involving the

2.

Sums of Powers of Roots.

309

Determine a function of the differences whose degree in the

coefficients is

four,

and whose weight

is six.

Assume

whence
D<j>

(6-4

i)rt

0203

(6-B

+ 3JE+ 2J9)0oi2 2 +

(3(7

+ 4D)

ai

<? 2

0.

3(7+

6 also 3(7 + IE = 0, giving (7=4; and Now let -4=1, whence .E 4D = 0, giving D = - 3; and from 65 + 3^ + 2D = 0, we have finally # = 4.
;

= -

Hence

Compare

Art. 42,

where the value of $

is

given in terms of the roots.

139.

Operation involving the .Sum* of the Powers of

the Roots.

Theorem.
(ttl,
tto,
3,

If

... O n )

= F(8l,

St y S3 y

.. Sr )

(1)

be any equation connecting a function of the sums of the powers with another symmetric function of the roots, we have then the
differential equation

d<t>

d$ ~ + aa
ai

-f- + -j- +
2

d$

a3

d<j> = + -r-

dF
s<>

an

aSi

T~ +

dF
2s i

T" + 3s^ as 3-+ as>

3
;

dF

+ r* r,!

dF
.

ds,-

For,

let

the roots be increased

by h

and comparing the

(1),

coefficients of

h on both sides of the equation

when

are substituted for

Si,

s2 ,

sr ,

we have

the required relation.

Employing the results of the last Article, we have, therefore, the following equation of operation connecting the coefficients and the sums of the powers of the roots
:

D=
D

tfo'f So -T-

+ 2!

+ 3s2

+ ...+ r

r_i

where

represents the result of substituting

for a in the

operator D.

310

Symmetric Functions of
this it follows that if /(tf
i, 2,
)
>

the Hoots.

From

i,

#2,

<*

n ) is

a function

of the differences, /(<,,

is

a function of the diffe#2,

*,

rences also

for
-

it

is

A/(*o>
since

*i> *2,
s

*n)

plain that when Df ( , # u = 0, and therefore Z>/(s , i,

#n)

=
=

0, 0,

*)

= - a^D.
4ai3+ 322 =
is also

Ex.

1.

-f

is

a function of the differences,

whence

4sisa

+ 32 2

a function of the differences.

Ex.

2.

=
2

/,

when

similarly transformed, gives

which

is

therefore a function of the differences.

MISCELLANEOUS EXAMPLES.
by squaring the determinant
a, )8,

1.

Prove,

of

Example

10, Art. 112, the following

:

relation

between the roots

Si

7,

8,

of the biquadratic

Sz

83

The student will have no difficulty in writing down for an equation of any degree the corresponding determinant in terms of the sums of the powers of the roots which is equal to the product of the squares of the differences.
2.

Prove, for the general equation,

so

Si

S2

This appears by squaring the array

11111.
(See Art. 123.)

)8

Miscellaneous Examples.
3.

Prove similarly,

for

the general equation,

By

the process of Art. 123, a series of relations of this kind can be established

and when the number of rows in the array becomes equal

to the degree of the

equation, the value of the determinant is the product of the squares of the differences, "When the number of rows exceeds the degree of the equation the as in Ex. 1.

For example, the value of the value of the corresponding determinant vanishes. determinant of Ex. 1 is zero for equations of the second and third degrees.
4. Prove, by means of the equations of Art. 130, that the sums of the powers can be expressed in terms of the coefficients, or vice versa, in the form of determi-

nants, as follows

Pi
Pi
1

2pz

pi
3/>3

pz

Pi

Si

6j3 3

=-

,,

.V!

2
*l

Sz

Si

S3

Sz

312
6.

Symmetric Func tions of

Prove, for the general equation,
So
Si

the Roots.

S2

S3

Sz

83

$4

Ss

x*

a*

Multiplying the two arrays

1
1

x-a
-,

x-&
(x

a
cr

7 7
2

a(x
a2

a)

)8)

&
that

(*-)
S2

P(x-P)

we show

is

equal to
S.)X
Si

SiX-Sz
SZ X

sz

x-

83

which

is easily

It appears in like

transformed into the proposed determinant. manner in general that the determinant of similar form'of

order p

-f 1 is

contains

factors of the original equation multiplied

equal to the corresponding symmetric function, each of whose terms by the product of the squared

differences of the
7.

roots involved therein.

coefficients of

Prove that the leading

1

Sturm's functions

(i.e.

/(#), /'(#)>

and

remainders) differ determinants

:

the n

by

positive factors only

from the following

series of

Miscellaneous Examples.
Assuming, therefore, for Rj and Aj the forms
Hj

313

2 r\x + rzx

+
+
0,

2 Aj = Ao + Ai x + A 2 z +

and substituting
Ao

in (1)

any root a of the equation /(x) =

. .
.

we have
a)
similar substitutions

Aio

+ A 2 cr +
a,

+
a->-

Ay-ia-'2
,

= /
;

Multiplying by of the other roots

relations of
;

a 2 , ...

a*- 1 , in succession

making
:

and adding the equations thus derived, we obtain by aid of the

the following system of equations
AI*I

Ex.

4, p. 296,

AO*O

~t~

+
+

A_/-2 Sj-t

~t~

A_/-i Sj-i

= =

0, 0,

Aol +

Al*2

Aj-2 Sj-l

\j-l Sj

0,

From

these equations

we

have, without difficulty,

o
Sj-l

si

S J-1 S J s j
,

4=V
Sj-2 Sj-l

*y-2
1

the value of

7,-

being so far arbitrary.

It appears therefore that the coefficient of

this multiplier only

the highest power of x in Rj differs (o*2*4 .... sy-z).

by

from the determinant

purpose we make the functions R and

proceed to show that the sign of 7,- is positive. For this use of the following relation connecting the successive values of

We

AM
To prove
this
;

MA
Jtk,

^f(x).

(2)

substituting for Itk+i

Rk-\ their values in terms of

A and

in the relation J2* +1

= QkRk -

k-i,

we
-i,

derive

Ak+i = QkA k -

Ak

B k +i
relations connecting the successive

by

aid of

which we readily obtain the following

:

functions

A
in which

Ri =f(x),

^1= /'(a,-),

=f(x).
coefficients of the highest

Now, comparing the

*" occurs only in

powers of a; in

(2)

observing that

Ak + i Rk, and making use of

the determinant forms above obtained,

*2*

we have
7* (^0*2*4
2)

1,

314

Symmetric Functions of

the Roots.

Also, calculating the value of R-> in the ordinary manner,

S
Si

we

easily find

whence

it is

seen that the value of 73

is

It follows,

of 7, that 73, 74, ...

from the relation just established between any two successive values and therefore, finally, that &c., are all positive squares 7,-,
;

rn -j, the coefficient of the highest power of x in

.

Jfy,

has the same sign as the deter-

minant (o *2 54 Sy-z) It may be observed that by aid of the preceding example the value of the quotient of AJ by T/ may be written as a symmetric function involving the roots and the
variable.

For example, when./

4,

we have

8.

Determine <i,

<f>z,

... </,

<pp

from the equations

Ans.

<f>j

is

given as a function of the

i
Oj

p
.

th

l}

degree in

0/

by the equation

e?
s2

e^
Sp.i

&
To

si

where
9.

sk

=
a,

6i

62*

03*

Op *.

If

)3,

7, 5

be the roots of the equation

#4

0,

calculate in terms of #o>

ao
6

-ff,

/,

/"

the value of the symmetric function

2
2

2(3a-)8-7-8) (3j3-7-5-a) (37-S-a-j8)

ao
6

2
.

Here
where
zi, 23, 23, 24

2=

4 5zi

2 2 zz z 3 ,

are the roots of the equation

a

GiTz2

+ 40* +

^-

3S" 2

0.

(See Art. 37.)

Hence, by Ex.

2, Art.

136,

Ans.

^-IKt + aJHI

Miscellaneous Examples.
10.

315

Prove that

n=
where

*o

6
()8

- y)~ (7 - a) 2 (a - 0) 2 (a - 8) 2

(j8

- S) 2 (7 -

5)

= IP + ,nJ*,

m=-

111.

use of a proposition which will be proved subsequently, namely, that any even, rational, and integral symmetric function of the roots, of the order 37,

The weight of "We now make

this function of the roots is 12,

and the order

6.

and involving the differences only of the roots, is, when multiplied by and integral function of #0, H, I, / (Compare Ex. 17, p. 124.)
o> -ff,

tfo

37
,

a rational

Hence, expressing the function whose order /, /, it is easy to see from the table

is 6,

and weight

12, in terms of

316
metric function
of the equation

Symmetric Functions of

the Roots.

by actually substituting tbe roots, and then comparing both sides when H, I, J are replaced by their values calculated from the

numerical

coefficients.

4 First we take the biquadratic equation 6z

Qx z

= 0, whose roots
/=!.

are 0, 0, 1,

whence

2=
Substituting in equation

8,

JT=-6,
we have

7=3,

(1),

Proceeding in the same

4
r

way with
5

the biquadratic equation

roots are

6z2

0,

whose

+ -v/% +

1,

we

find

2 =
whence
and

768,

JET=-1,

J=8,
21

/=-4;

192

+ m,

= - 2x192,

m = 3x192;

and, finally,

12. Calculate the determinant

SO

si

*s

*2

*3

*4

in terms of the coefficients of a quartic.

This determinant is a function of the differences of the roots

(see

Ex.

2, Art.
it

139)
;

we may
if

therefore remove the second term of the quartic before calculating

and

the equation so transformed be

4
2/

+ P2y2 + Pay +

P4 = 0,

- 2P2
- 2P2 - 3P3

4 { 8P2

P4 - 2P2 * - 9P3

- 2P2 - 3P3

2P2 - 4P4
2

Substituting for

2,

PS,

P4

these values,

we have

the same result as in the preceding example. (Compare Ex. 13. If be the roots of the equation o, &, 7, 8
-f

3, p.

311.)

6a 2 a; 2

-f

-t-

0,

Miscellaneous Examples.
express

317

It J
,

t,

G,
4

of the equation
4*i a:
3

Jo*
in terms of
ff,

Gszx*

4* 8 #

84,

= 2(* + =
485-2

o)

=
2

I, /, G.

-4*.
and by the aid of the relations

H
o

H
o
2
T,

I,
r
*o

^-= 2

-T 4 o o

/
,

~~*~~"~~* 3 o <V

O _
t

G*

45-3

^(sx _

ffo

/),

&2 +

4tf,

SQ

(a ; 7t _
s

/ =
14.

_T
o*

QO

When^?

is

even, prove that

2(ai
Since

at)P

= to8p -psiSp-i + $p(p~

1) *2*P -2

-&
-^*

- &c.

changing

a;

into

ai, 02, 03,

... a,,

in succession, and adding the results on both

sides of the equations thus obtained,

we

find
\

22(01

tt2)P= *0*p

p*\*p-\

p p
.

S2Sp-2

JP*1 *p-l

0*i,

where
term.

all

the terms on the right side of this equation are repeated except the middle

Thus

2 (ai -

a2 ) 4

so *4

- 4i s 3 +

3*2 2 ,

(Compare Ex.
2
,

Art. 139.)

&c.
^>'(

15.
o,
)8,

Form

the equation whose roots are

</>'(<*),

#'(0), ^>'(y),

5 )>

where

7, 8 are the roots of the equation

<p (x)

^o^

4#i

x3

60 2 %z

403

0.

-27/ 2 )

o.

a
16.

If

2 (a -

0)

(0

-7

2
)

(7

a)

(x

5)*,

when expanded, becomes

EQX* + 4Kix3 +
prove that
JToaj37

6 jfiT2 x*

JTi

(07

+ 7 + 0) + ^2 (a + 08-7)(7-)(a-/3) A = I3 - 27/ 2
.

7)

_+

16

where
17.

Prove that
ao*2()8

+7 a

5)

()8

?)

S)

192

(3

/- 2JT/).
36 a Q

18.

Prove that

5)4 (3

7)2

8)8

512 (a<?l

HJ +

CHAPTER

XIII.

ELIMINATION.
Being given a system of n equations, n between or non-homogeneous between variables, homogeneous n-I variables, if we combine these equations in such a manner as to eliminate the variables, and obtain an equation It = 0,
containing only the coefficients of the equations the quantity in a rational and integral form, called their is, when expressed
;

140. Definitions.

Resultant or Eliminant.

In what follows we

shall be chiefly concerned

with the dis-

cussion of two equations involving one unknown quantity only. = asserts that the two In this case the equation equations are consistent that is, they are both satisfied by a common

value of the variable.

elimination

may

now proceed to show how the be performed so as to obtain the quantity R,

We

illustrating the different methods by simple examples. Let it be required to eliminate x between the equations
ax*

+ 2bx +

0,

a'x*

+ 2b'x +

c'

0.

Solving these equations, and equating the values of x so obtained, the result of elimination appears in the irrational form

Multiplying by aa'

we
a'b

obtain

aV Squaring both
vanish

= a*/b'*-a'c' -

sides,

when

aa' vanishes),
(ac

and dividing by a a? (for R does not and then squaring again, we find

R=

V) (a'c'~

b'*}

(ac

a'c

2bb')\

Elimination by Symmetric Functions.

This method of forming the resultant
p.ication, as
it is

319

very limited in ap-

not, in general, possible to express by an aljebraic formula a root of an equation higher than the fourth Other methods have consequently been devised for decree.
is

de:ermining the resultant without

first solving the equations. of elimination by symthe method explain metric functions of the roots of the equations. 141. .Elimination by Symmetric Functions. Let two

We now proceed to

algebraic equations of the

(x)

th

and n th degrees be
. . .

= a Q x m + aixm ~ l + a,x m - z +
&o#
M

+ am =
+
bn

0,

^ (x) =
and
-the

n~l
t>ix

M
.
.

n ~2

=
;

let it be required to find the condition that these equations should have a common root. For this purpose let the roots of

equation 0(a?) have a common root

be
it is

ai,

a2,

a.
and

necessary

If the given equations sufficient that one of the

quantities
^(QJ),
i//(a 2 ),

...,

\fr(a m )

should be zero,

or, in

other words, that the product

\f;(a 2 ) if

^(a,)

(a s ) ..

i//(a

m)

now, we transform this product into a rational and integral function of the coefficients, which is always possible = 0, as it is a symmetric function of the roots of the equation (x)
should vanish.
If,

we

shall

be the roots of the equation

have the resultant required. Further, = 0, we have (x)

;//

if j3i, /3 2 ,

|3 n

(a 2

|3 n ),

*(a) - b(a m If

/3i)

(a m

/3 2 )

(a m

(3 n ).

n factors, and multiply these signs of the equations, taking together the factors which are situated in the same column, we find

we change the

<C * (,) * (o.)

* (a.)

(- 1)""' 4

((3.)

* (ft)

. .

f (ft,).

320

Elimination.

We may therefore take

for both these values of
ficients of
i//

R
,

are integral functions of the coef-

(#)

which vanish only when (x) a id (x) (#) have a common factor, and which become identical wh3n

and

t/

they are expressed in terms of the

142.

coefficients.
(1).

Properties of the Resultant.

The order of

the resultant of two equations in the coefficients is equal to the sum of the degrees of the equations, the coefficients of the first equation
in the degree of the second, and the coefficients of the entering second entering in the degree of the first.

This appears by reviewing the two forms of R in (1), th for in the first form a dm enter in the n i, th b n enter in the m degree, and in the second form b Q fti,
Art. 141
;
, ,
. .

may be seen that two terms, one selected from degree. each form, are (- l) mn b m a mn and a n b nm (2). If the roots of both equations be multiplied by the same
Also
it
.

quantity p, the resultant

is

multiplied by p

mn
.

This
op

j3 g

any one of the mn factors of the form becomes p (ap - )3 ? ), and therefore p mn divides the resultis

evident, since

ant.
is

From

this

we may conclude

that the weight of the resultant

mn, in which form this proposition is often stated. (3). If the roots of both equations be increased by the same
is

quantity, the resultant of the equations so transformed

the resultant of the original equations.

equal to

For we have

R=a bn(ap -p
where

q ),

signifies the
;

form ap same increment.

(4).

- p q

continued product of the mn terms of the and this is unaltered when ap and f3 q receive the

If the
when

roots be changed into their reciprocals, the value of

R obtainedfrom the transformed equations remains unaltered, except

in sign

mn

is

an odd number.

Making

this transformation in

Properties of the Resultant.

321

we have

but

!,...-(- l)"^', 00
substituting,

ftft

= (-

1)

[ Co

we

obtain
fl

R=
e. g.
,

(- !)'

n (a, - ft)

= (- !)"".

From this it follows that in the resultant of two equations the coefficients with complementary suffixes of both equations,
(i

tf

i,

0m-i, &c.,

may

be

all

interchanged without alter-

ing the value of the resultant.

(5).

If loth equations
that
is,

mation

be transformed by homographic transforby substituting for x

\X +

jl

and each simple factor multiplied by X'x + p, to render the new R' = (Xp - A'/ equations integral; then the neic resultant

To prove

this,

we have

$
also

(x)

=a (xQ

ai) (x

at )

(x

a m ),

i / \ P*1 * M \' \ ( - A x - ^ x - a r becomes (A f a,.) ^a A A

Multiplying together
a
I,

all

the factors of each equation,

A'a 2 ) ... (A

becomes a
^o

(X

AX) (A A'fr) (A

A'a m ),
A'j3 n )

(A

\%)

... (A

Also, since a r ,

/3 r

are transformed into

0r

^ A

A/ A ar

^,

^ A

A p,-

x/

^,

ar

ft becomes

(Xft

322
whence
a nb

Elimination.

mU

(or

j3 r )

becomes a

m
(A//

AV)

wn

II (a r

]3 r ),

that

is,

the resultant calculated from the

new forms

of

(x)

and

iff(x) is

(\n'-\'n)

mn K.
;

This proposition includes the three foregoing

collectively equivalent to the present proposition.

and they are

Method of Elimination. When two = 0, of the m th and n th = 0, and \[/(x) equations 0(#) degrees respectively, have any common root 0, we may assume
143. Euler's

+(*)

(a

0)

M*)
.

where
0i (a)

^Pi%

m- 1 + 2 xm~* + p

+p

the coefficients heing undetermined constants depending on

0.

Whence we have

Now, equating the coefficients of the different powers of x on both sides of the homogeneous equations of the first equation, we have

an identical equation of the (m +

n- l) th degree.

m+n

and ; degree in the eliminating these constants by the method of Art. 125, we obtain the resultant of the two given equations in the form of
, ,

+ n constants p l9 p^ ... p m q ly qz ... qn

a determinant.
EXAMPLE.
\
Suppose the two equations

ax2
to have a

bz

0,

aix~

bix

c\

common

root.

"We have identically

or
(q\a

-p\ a

0.

Sylvester's Dialytic
Equating to zero
geneous equations
q\a
q\b
all

Method of Elimina tion.

we have

'

'2

>

the coefficients of this equation,

the four homo-

-p\(ii

0,
0,

02
q-ib

-p\b\ - p*(ii =
-p\c\ -p*bi

q\c

=
=

0,

q<>c

q\ t qz,

pzci

and eliminating the constants

p\,

jt> 2 ,

we

obtain the resultant as follows

a
b
c

0i

a
b
c

b\

a\

= o.
c\

li

a
same as that of Art. 140.

The student can

easily verify that this result is the

144. Sylvester's Dialytic

Method of Elimination.
;

This method leads to the same determinants for resultants as

the method of Euler just explained but it has an advantage over Euler's method in point of generality, since it can often be

applied to form the resultant of equations involving several

variables.

Suppose we require the resultant of the two equations

(x)

= a xm +
n

tfiz*"1

+ a z a^- z +
b2 x
n ~*

+ am =

0,

$(x) = b x

+ hx*- +

+... +

&=

0,

we multiply

the

first

by the

successive powers of x,

and the second by

thus obtaining m + n - 1.

We

+ n equations, the highest power of x being have, consequently, equations enough from
^
*
'

which

to eliminate
t
>

considered as distinct variables.

Y2

324

Elimination.

EXAMPLE.
In the case of two quadratic equations
ax*

bx

0,

a\ x*

b\

x +
0,

c\

0,

we have

=
1

+ bx +

0,

+ ax

=0,
0;

from which, eliminating x now replacing rows

:

3
,

z
,

x,

we
b

get the same determinant as before, columns

145.

Bezowt's Method of Elimination.

will

The general
it

method
first

be most easily comprehended by applying

in the

instance to particular cases. proceed to this application the are same degree, and (2) when of the (1) equations when they are of different degrees.
(1).

We

Let us take the two cubic equations

ax*

bx*

+ ex + d =

0,

#i#

+ ha? + c& + d =
L

0.

Multiplying these two equations successively by

ai

and

a,

a\x +
aiX*

bi

ax +
ax*

b,

+ biX +

+ bx +

c,

and subtracting each time the products

three following equations
(ah) a?
(acjx*
:

so formed,

we

find the

+(ac }x
l

+ (ad^ =
(bd,)

0,

{(ad!)

+ (bd}}x +

0,
0.

Bezoufs Method of Elimination.

325

eliminating from these equations ar, x, as distinct variables, the resultant is obtained in the form of a symmetrical

By

determinant as follows

(ad)
(a Cl )
(ad,) (ad,)

(ad,)
(bd,)
(ed,)

(be,)

(bd,)

To render

the law of formation of the resultant more apis

parent, the following mode of procedure Let the two equations be

ax*

given

bo?
x*

+
+

ex*
2

+ dx +
+ d,x +

0,

<?i#

61

whence, following Cauchy's mode of presenting Bezout's method, we have the system of equations
a bx3 + ex2 + dx +
e

+
ax + b
,x +
bi

c,x

ex-

+ dx + + diX +
dx +
,x +

e
e\

CiX* c
Ci

ax2 + bx +

e
e\

+
ax*

ix
bx*

-t-

ex +

which,

x3 ,

a?,

of integral, lead, on the elimination resultant the the for to form x, following
:

when rendered

(ab,)
(ac,)

(ac,)

(ad,)
(ae,)
(be,)

(aefi
(be,)
(ce,)

(ad,)
(ae,)

+ +

(be,)

+
+

(bd,)
(cd,)

(ad,)
(ac,)

(bd,)

(be,)

(ce,)

(de,)

326
If,

Elimination.

now, we consider the two symmetrical determinants,

(obi)
(ac,)
(ac,)

(ad,)
(ae,) (be,)
(ce,)

(ae,)
(be,)
'

(ad,)
(ae,) (be,)

(be,)

.(bd,)
(cd,)

(ad,)
(ae,)

(ce,)

(bd,)

(de,)

the formation of which

is

obtained

by adding

is at once apparent, we observe that It the constituents of the second to the four
first.

central constituents of the

Similarly in the case of the two equations of the fifth degree

ax 5 + bx* +

ex*

dx*

+ ex + / =
,=

0,
0,

the resultant

is

obtained from the three following determi-

nants
(ab,)
(ac,)

(ac,)

(ad,)
(ae,) (af,)
(bf,)

(ae,)
(af,)

(a/,)
(If,)
(of,)

(ad,)
(ae,)
(af,)

(be,)

(bd,)
(be,)
(ce,)

(be,)
(ce,)

(ad,)
(ae,)

(bf,)
(of,)

(bd,)
(be,)

(df,)

(de,)

by adding the

constituents of the second to the nine central

first,

constituents of the

and then adding the third

to the central

constituent of the determinant so formed.

The student

will

have no

difficulty in applying a similar process of superposition to the formation of the determinant in general.
(2.)

We
ax*

take

now

the case of two equations of different

dimensions, for example,

bx*

+ ex2 + dx

+-

0,

a,x~

+ b,x +

c,

0.

Multiplying these equations successively by

a,

and

ax2 3
(ax

a,x +

b,

+ b)x z

Bezoutfs Method of Elimination.

327

and subtracting each time the products two following equations

:

so formed,

we

find the

+
(aci)x*

(aci)ar

- da^x -

ea^

0,
eb^

{(bd)

da^x-

{dbi

+ ca^x -

0.

If,

now, we join to these the two equations

a t x 3 + biX* + CiX = 0,

+ biX +

0,

have four equations by means of which * 3 #2 x can be eliminated whence we obtain the resultant in the form of a

we

shall

determinant as follows
(ffbtj

328
so that

Elimination.

we must be on our guard

lest

~ the factor a mm n

(i.

e.

the

result of substituting these roots in

(x))

enter the form of the

resultant obtained.

the above ease

From these two equations the (1) following n equations

:

we

derive, as in

+ a z xm

~2

+ am

am

a xn

~l
1

+ a xn
l

-' t

+
+

a^
b n_ l

_ ~

a n xm n +

a^x-" +
1

+ am
'

l^x"- +

b 1 x^~+ ...

b n x"

ih integral, are all of the (m l) degree ~l ~z m m x as independent quantities whence, eliminating x , x between these n and the m - n equations,

which,

when rendered

'x

m-2 + m~ + b,x
l

b2 x

m~ z +

=0,

bQ x

+ bi%n

-1

bn

0,

we

obtain the resultant in the form of a determinant of the

th

order, the coefficients of the first equation entering in the degree n, and the coefficients of the second equation entering in the

degree

whence

it

and

that the resultant as obtained

appears that no extraneous factor can enter ; by this method has not been
of the zero roots.
(x)

affected

by the introduction

If

R be the

resultant of two equations,

to

0,

^ (x) =
f

0,

whose degrees are both equal

system

m, the resultant

of the

X0

(x)

+ nt (*) -

0,

X> (x)

(x)

for each of the

minors

(a r b s ) ,

which in Bezout's method con-

Bezoufs Method of Elimination.

stitute the

329

determinant form of

J2,

becomes in

this case

Aa r +

inb r ,

\'a r \'a s

f/b r

whence
order m.
146.

\a s + pb s R' = (A// ,

+ f/b s
-ft,

m
A'ju)

since

is

a determinant of

We

count of a method which

conclude the subject of Elimination with an acis often employed, but which has the

disadvantage, when applied to equations of higher degree than the second, of giving the resultant multiplied by extraneous

The process about to be explained is virtually equivalent to that usually described as the method of the greatest
factors.

common
equations

measure.
this

In forming by
ax~

method the resultant

c

of the

two quadratic

+ bx +

0,

2 #i# + biX +

Ci

0,

we multiply

these equations successively

!

and

tf,

Ci

by and

c,

and subtract the products

linear equations
(abi)

so formed.

We
0,
;

thus find the two

x + (ad) = = (aci) x + (bci)

from which, eliminating

x,

we have

it

As the degree of this expression is four, and its weight four, can contain no extraneous factor, and is a correct form for the
To form by
the same process the resultant of the cubic

resultant.

equations

a& +
we

bx*

+ ex + d =

0,

a^y?

b^ +

CiX

d^

=
a,

0,
di

and subtract each time the products

(ab,}x-

multiply these equations successively by a^ and so formed.

and

d,

We have then

(ac,}x

(ad,)

0,

330
Now,

Elimination.
eliminating x between these two quadratics

by means

of the formula above obtained,

(ab,)

we

find for their resultant

(ad,)

(ac,)

(ad)
(bd,)

(ad,)

(ad,)

(cd,)

(ad.)

(Id,)

(cd,)

an expression whose degree is 8 and weight 12, in place of degree 6 and weight 9 whence it appears that it ought to be divisible by a factor whose degree is 2 and weight 3. This factor must therefore be of the form I (be,) +m(ad,). We proceed
;

now

to

show that

it is (ad,)

and

to find the quotient

when

this

factor is removed.

For

this purpose, retaining only the

terms which do not

directly involve (ad,),

we have
+
(ca,)(bd,)},

(ab,)(cd,){(ab,)(cd,)

which

is divisible

by

(ad,), since

(be,) (ad,)

(ca,) (bd,)

(ab,) (cd,)

0.

Expanding the determinants, and dividing

find ultimately the quotient
(ad,)*

off

by

(ad,)

we

- 2 (ab,)

(cd,) (ad,)

(bd,) (ca,) (ad,)

K) K) + (ab,) (bd,)*
3

(ab,) (be,) (cd,),

which, being of the proper degree and weight, is the resultant. If we proceed in a similar manner to form the resultant of

two biquadratic equations, by reducing the process to an elimination between two cubic equations, we shall have to remove an extraneous factor of the fourth degree, which is the condition
that these cubics should have a
ratics

common factor when the biquad-

mon

from which they are derived have not necessarily a comand in general, if we seek by this method the factor
;

resultant of two equations of the n th degree, eliminating between th two equations of the (n l) degree, we shall have to remove

an extraneous factor of the order 2n

fore is inferior to all the preceding

4.

This method there;

methods

and

it

cannot be

Discriminants.

331

conveniently used except "when, from the nature of the investigation, extraneous factors can be easily removed.

The discriminant of an equation inis a variable the simplest function of the coeffivolving single in a rational and cients, integral form, whose vanishing expresses the condition for equal roots. have had examples of such
147. Discriminants.

proceed to show that they come under eliminants as particular cases. If an equation = has a double root, this root must occur once in the f(r) and subtracting xf(x] from nf(x), the same equation ./"(;/') = root must occur in the equation nf(x] - xf(x) = 0. This is an equation of the (n - l) th degree in x and by eli= minating x between it and the equation/' (x] 0, which is also of the (n - l) th degree, we obtain a function of the coefficients whose
functions in Arts. 43 and 68.
; ;

We We

vanishing expresses the condition for equal roots. The degree and its of this eliminant in the coefficients of f(x) is 2 (n - 1) - 1), as the is be seen specimen by examining weight n(n may
;

terms given in section

(1),

Art. 142.

Expressed as a symmetric

function of the roots of the given equation, the discriminant will be the product of all the differences in the lowest power which

can be expressed in a rational form in terms of the

coefficients.
2

the product of the squares of the differences n (ai - a 2 ) can be so expressed; and since it is of the 2(n - l) th degree in

Now

any one
is

degree in all the roots, it follows that the discriminant multiplied by a numerical factor
root,

and
~
)

of the

n(n

tk

l)

z n l n (a - a 2 ) 3 and is, moreover, identical with equal to a the eliminant just obtained. If the function /(#) be made homogeneous by the introduc(

tion of a second variable y, the

two functions whose resultant

is

the discriminant of f(x] are the differential coefficients of f(x) with regard to x and y, respectively. In the same way, in general, the discriminant of a function

homogeneous in any num-

the result of eliminating the variables from the n equations obtained by differentiating with regard to each variable in turn.
is

ber n of variables

332

Elimination.

EXAMPLES.
1
.

Find the discriminant of

0o# 3

30i a;

3#2#

03

0.

"We have here to find the eliminant of the two equations

aox2
01 #
2
-f

20i #

-f

02
03

0,

202 X

0.

This

is,

by

Art. 140,

or

it

may be

written in the form of a determinant, as follows, by Art. 144

20i
3

'

0.

It can be easily yerified that this value of the discriminant is the

same

as that

already obtained in Art. 42.

2.

Express as a determinant the discriminant of the biquadratic

o 4 4-

4i

a;

6rt 2 #

+ 40 3 # +

0.

"We have here to eliminate x from the equations

a Q xz
aix*

+
+

3aiz2
3a 2

+
+

+
+

0,

xz

Za^x
is

a*

=0.

By

the method of Art. 144 the result

00

Determination of a Root common

4.

to

two Equations.

333

roots of the biquadratic of

Prove by elimination that Ex. 2.

J=

is

one condition for the equality of three

Since the triple root must be a double root of

and therefore a single root of

single root of
ZT-j

a\x-

2a 2 x

+
a2

as

=
0,

and since

it

must

also be

a^x*

2ai x

it

follows from the identity

that the triple root must be a root

common
2

to the three equations

aox

2a\x
2a 2 x

+ +

a2
3

0,

a^x* a2 x*

= =

0,

+ 23* +
#1

04

0.

Hence the condition

flO

#2 #3

ai

a?

That the other condition for a

p.

triple root is

/=

may

315

for

when three

roots are equal the discriminant

be inferred from Ex. 10, must vanish, and it is of the

the product of

form // 3 +
5.

w/ 2

Prove that the discriminant of the product of two functions

is

their discriminants multiplied

by the square

of their eliminant.
;

This appears by applying the results of Art. 141 and the present Article for the product of the squares or the differences of all the roots is made up of the product of the squares of the differences of the roots of each equation separately, and the
square of the product of the differences formed by taking each root of one equation with all the roots of the other.

148.

Determination of a Root
If

common

to

two

Equations.

be the resultant of two equations

U = a m y^
and
a

any common

root,

then

dR
da
i

dR
do, 2

dR
dn^

da

334

Elimination.

and 6 a Q - U and To prove this we substitute in R, for # F, and obtain an identical equation connecting U, V which is satisfied by every value of x, and which is of the form
,

R=
whence

Iffy

Fi/>

dR
ddp

Tr deb

T_

(l\l>

dap

dcip

and when a

is

common

we have,

substituting this value for

= 0, and root of the equations 0, x in the preceding equations,

F=

dR
a

dR
dap+l
9

dap

which proves the proposition. double root of an equation can be determined in a similar

manner by
149.

differentiating the discriminant A.

Symmetric Functions of the Roots of two

If
it
a, , ,
. .

be required to calculate a symmetric funcaw of the equation tion involving the roots a a 2 a 3 ~ -z m m m = a x + a,x + + a m = 0, + a2 x (x) (1)

Equations.

()

along with the roots

^(y) -

)3i, |32, j3 3 ,
l

j3,

of the equation

n~ n ~z n ^y + ^y + ^y +
:

o,

(2)

we

proceed as follows

Assume a new
equation

variable

connected with x and y by the

;

= \x + ny

and

let

ybe eliminated by means

an
equation of the
th

result is

volve X, ju, and t in the n any of the preceding methods from this equation and (1). th obtain an equation of the mn degree in t, whose roots are the
Jim values of the expression Xa +
/u/3.

of this equation from (2). The n th degree in x whose coefficients inpower. Now let x be eliminated by

We

Miscellaneous Examples.
If,

335

now,

it
\f/

and (.*) form the sum of the (p + q) ih powers of the roots of the equation
of
,

be required to calculate in terms of the coefficients ? we (y) any symmetric function such as 2a^ |3

in

t.

We
fj..

thus find the value of

S (Aa

+ ju/3)^ expressed in

terms of the original coefficients and the several powers of \

and

The

coefficient of

\p ^
p
/3'
7

the required value of 2a

in this expression will furnish in terms of the coefficients of

j(x) and^(y).
MISCELLANEOUS EXAMPLES.
1.

Eliminate x from the equations

ax-

bx

0,

x*=l.
Multiplying the
first

equation

hy

x,

we
+

have, since 3?
a

1,

bx~

ex

and multiplying again by

x,

we have
ex2

ax + b

0.

Eliminating x~ and x linearly from these three equations, the result

as a determinant
'

is

expressed

If the

method of symmetric functions

(Art. 141)

the second equation substituted in the

(a
2.

first,

the resultant

be employed, and the roots of is obtained in the form

ba?

c)

(aar

bw

c)

(aw

c).

Eliminate similarly x from the equations

ax*

bx3

cx~

dx +

e
5

=
=

0,
1.

The

result is a circulant of the fifth order obtained

by a

process similar to that

five factors

of the last example. can be written down

By
(cf.

aid of the

method of symmetric functions the

Ex. 27,

p. 284).

An

analogous process

may

be applied in

general to two equations of this kind. 3. Apply the method of Art. 143 to find the conditions that the two cubics
<j>

(x) (x)

ax3
a'x*

+
+

bx2
b'x*

ex

+d =

0,

iff

+ c'x+d'=Q

should have two

common

roots.

336
"When

Elimination.
this is the case, identical results
ty

must be obtained by multiplying

</>

(x)

by

the third factor of

(x)

and

i//

(x)

by the

third factor of

$ (x).

We have,

therefore,

(\'x

+ p) $ (x} =

(xx

At)

$(x),

where

A,

/i,

A',

/*'

are indeterminate quantities.

This identity leads to the equations

\'a
\'b
\'c

fj.'a

- \a - \b' r

=
fj.a'

0, 0,

= + pb Ac' - \d' - p.c' = \'d + - /j.d' = n'd

/jib' (j.'c

0,
0,
0.

Eliminating

A.', /*',

A,

p from every four

of these,

we

obtain five determinants,

is

whose vanishing expresses the required

conditions.

There

a convenient notation

in use to express the result of eliminating from a number of equations of this kind. In the present instance the vanishing of the five determinants is expressed as
follows
:

0,

the determinants being formed

4.

by omitting each column

in turn.

Prove the identity

a2 a
a'
2

2a)3
oj8' -f

&
P
j8'

2o'j8'

This appears by eliminating x and y from the equations

ax
for

+ py =
derive
(ax

0,

ax +

ffy

from these equations

(ax

we
0,

)8y)

+
is

0y) (a'x

+ py) =

0,

(a'x

j3

V)

0.

The determinant above written

latter equations
;

the result of eliminating #2 , xy, and y 2 from the and this result must be a power of the determinant derived by
linear equations.

.eliminating x,
5.

y from the

Prove similarly

2oj8j8'

s
3a' 2 j8'
2

(a)3'

a'

3a')8'

Miscellaneous Examples.

337

This appears by deriving from the linear equations the following equations of the third degree
:

(ax

-f

0y)

0,

(ox
3
.

ft/)

(ax

-f /8'y)

0,

&c.,

and eliminating
6.

x3 xz y,
,

xy*,

Prove the result of Ex. 12,

p. 278,

by eliminating the

constants A, p,

A.',

n',

from four equations

,

Act

4-

fJ.

Aj8 -f

fj.

connecting the variables in homographic transformation.

7.

Given

U = Auz +

2Buv +

Cv*,

2bxy

determine the resultant of

Since

U and V considered as
~U= A(u - av)(u-

functions of x, y.

fiv),

if

U and V vanish

for
;

common

values of x, y, some pair of factors, as u

av and

whence forming the resultant of u presenting the resultant of u and v by JS(w, v), we have
av, must vanish

av and u

a'v,

and re-

R(u
and multiplying
all

av,

a'v)

(a

a')

R(u,

v)

',

these resultants together,

we
(a

find
8
jB')
(ft

E(UZy
or

r.)

= A*A'*(a -

a')

()8

0')

a')

{*(,

v)}

4
,

8.

equation f(x)

Prove that the equation whose roots are the differences of the roots of a given = may be obtained by eliminating x from the equations

/(*)

0,

/'(*)

+/"(*)

+/'"(*)

Y^ +
44).

&c.

and determine the degree of the equation in y (cf. Art.

CHAPTER

XIV.

COVARIANTS AND INVARIANTS.

150. Definitions.

In

this

and the following Chapters the

.

notation
(0o, 0i,
2,

..

0)(a?,

yY

will be

employed to represent the quantic

~l

+ an yn

which
mial

is

a homogeneous function of x and y, written with binoIf

coefficients.

we put y =

1, this

quantic becomes

Z7" w

of

Art. 35.

Let

be a rational, integral, and homogeneous symmetric

37,
2
.

function, of the order

of the roots a 1? a 2 , a 3 ,
.
.

a n of the

equation TJn

(0

0i,

0) (#,

l)
;

0, this
if

function involv-

ing only the differences of the roots

1

then
1

1
'

- x
,
. .

'

a*.

x*

an

-X

a w respectively, the result multibe substituted for a 1? a2 w remove U n n if it infractions) is a covariant of (to plied by volves the variable x and an invariant if it does not involve x.
,

From
that

this definition of

an invariant we may
a2 a 3 ,
,
. .

infer at once

<V $
is

(ai,

a n)

is made up of a number of terms an invariant of Un when of the same type, each of which involves all the roots, and each root in the same degree OT.

Formation of Covariants and Invariants.

These definitions

339
<

may

be extended to the case where

(the

function of differences) involves symmetrically the roots of seve= 0, &c., the roots of these = 0, = 0, ral equations r p q

equations entering

in the orders w, w', w", &c.

a,

respectively.

We may substitute for each root

fractions

--- as before,
-x
*"
. . . .

and remove

&c. If the result the multiplier ~U/ II r obtain a of covariant the system of involves the variable a?, we

by

fU

P9 quantics of the system.

U V U
q,

r,

&c.

and

if it

does not,

is

an invariant

Formation of Covariants and Invariants. We to show how the foregoing transformations may now proceed
151.

be conveniently effected, and covariants and invariants calWith this object, let the culated in terms of the coefficients.

symmetric function of the differences of the roots be expressed

in terms of the coefficients as follows
00*0 (a i,
Cto,
.

a n)

= F(a 09

0i,

2,

0).

Now, changing the roots into their reciprocals, and consequently a into n &c., a r into a n _ r) &c. (that is, giving the suf,

fixes their

complementary values), we have = F(a n 0/_i, O/>(ai, ao, ... a w )

>

0o),

is an integral symmetric function of the roots, and i// This the corresponding value in terms of the coefficients. function is called the source* of the covariant derived therefrom.

where

Again, substituting

01

x,

a2

a?,

an

-x

for

cti,

az

...

a,,,

and consequently

r,

&o., for a ri &c. (see Art. 35),

we

find

Thus, by two steps

we

derive a covariant from a function of

its

the differences, and find at the same time

lated in terms of the coefficients.

equivalent calcu-

To

illustrate this

mode

of procedure

we

take the example in

the case of the cubic

2

tfo

S (a -

|3)

= 18 (a? - a a2 )

* This term

was introduced by Mr. Roberts.

z2

340

Covariants and Invariants.

a ly # 2 , a

into # 3 ,

whence, changing the roots into their reciprocals, and a a oy we have #2, i,

2a 3 (|3 - jY = 18 ( 22 Again, changing a, j3, 7 into a #, ]3 #, 7 #, and

2
ff

i,

ff 2

#3

into

Z7i,

II^

Z7" 3,
2

respectively,

we
2
)

find

*o S(/3

7)

(*

= 18

(tT,

ZTiZTi).

The second memher

Z7i Z7" 3

of this equation

becomes when expanded

a^ a 2 )

Uf =

(^0^2
is

a\] a? + (a as

+ (i #3 .

/).

This covariant
as 5^, since -5" is

called the Hessian of Z73

coefficient.

We

refer to it

As
quartic

leading a second example we take the following function of the

:

its

2 3 S(/3- 7 ) (a-S) = 24(

^-4^

3^2 );

(1)
2,

into a^ a 3 , # 2 ,

whence, changing the roots into their reciprocals, and a w Oi 9 we have i,

>

<%,

'S(7

)3)

(8

a)

= 24 (atao-^a, + 3<).
alter equation (1)
:

These transformations, therefore, do not

again, since in this case ^ (a, |3, 7, S) is a function of the differences of the roots, ty is unchanged when a - a?, j3 - a?, &c. , are substituted for a, j3, 7, S. infer that a<>a - 4aia z + 3# 22
.
. .

We
.

is

an invariant of the quartic

Z74

We

observe also, in accordance with what was stated in

Art. 150, since

that each of the three terms of which $ is made up involves all the roots in the degree CT, which is here equal to 2. In a similar manner it may be shown that

= - 432
is

an invariant of the quartic.

Properties of Covariants

and

Invariants.

341

There is no difficulty in determining in any particular case whether $ leads to an invariant or covariant, for if ^ leads to an = ;//, that is is unchanged (except in sign, when invariant,
its type term is the product of an odd number of differences of the roots, i. e. when its weight is odd), when for the roots their reciprocals are substituted, and fractions removed by the simplest

a From another point of view an multiplier (en a 2 a 3 ... a n ) invariant may be regarded as a covariant reduced to a single
.

term.
152.

Properties of Covariants and

is

Invariants.

a homogeneous function of the roots, the covariant derived from it may be written under the form

Since

U'
ai-#
where
J

a 2 -#'

an

-.r,

the order, and K the weight of $. Also, as $ is a function of the differences,

is

we may add
Q).

1 to

each constituent such as

X by
,

thus obtaining
x,

Of multiplying each constituent

Again,

the covariant becomes

a nx \ I /J a n - x)

U
which

w
fL

~~;T

a^ 9

(~"a&
\ai

-x

~ ~aiX
>

a*

>

may
(-

be reduced to the form

IYU X- ^
2

/J_
/

^_

T^l
a2 /

_!_ f<

'i__iV
1
\x
1

where
an

whence

it is

proved that the covariant form

-x
is

unaltered
;

when
i,

for a b
... a n

a-.,

... a

rt ,

a?,

their reciprocals are sub,

stituted

fl

2,

changed into a n

_i,

respectively,

and the

result multiplied

by

(- 1)"^*-

*.

342

Covariants
if

and Invariants.
is

Now
form

any covariant whose degree

l

m be

written in the

(S^S ,S,,...S.)(x,l)';
changing a
,

(1)

... a n ,

%,

into

an

a n. h ... a0j -,

we have

another form for this covariant, namely,

Cl, (72) ...

CL)

-,

and
(1),

as this

we

an integral function of x of the same type as have, by comparing the two forms,
form
is
,

m = ms - 2 K

= (-

1)" ft,

= (-!)" ft_.

thus determining the degree of the covariant in terms of the order and weight of the function $, and showing that the conjugate coefficients (i. e. those equally removed from the extremes) are
related in the following

way
.

If F(a m
(~l)
K
,

#1,

#2

... a n )
,
. .

be

any

coefficient

of the covariant,

of

F(a n </_!, an _2 From the expression TS and K, namely ms

:

tf

is its

conjugate.

for the degree of a covariant in terms

2c,

we may draw
n-sj

the following im-

portant inferences
(1).

If

a^

is

an invariant,

2*.

For, in this case $ and i/ are the same function, and conse- K also the same. quently their weights K and ms
(2).

All

the invariants of quantics of

odd degrees are of even

order.

zj

For if n be odd, it is plain from the equation must be even, and K a multiple of n.
(3).

nvr

2/c

that

All covariants of quantics of even degrees are of even

in this case nzj - 2c

degrees.

For
(4).

is

even.
is

The resultant of two covariants

always of an even degree

in the coefficients of the original quantic.

Examples.

343

orders

For, the degree of the resultant expressed in terms of the and weights of the covariants is
/

'

'\

'
i

/"

O\

Of

'

'

We add

some examples in

illustration of the principles ex-

plained in the preceding Articles.

EXAMPLES.
1.

2.
3.

Show Show

that the resultant of

two equations

is

that the discriminant of

any quantic

is

an invariant of the system. an invariant.

Prove directly that any function of the differences of the roots of the cova-

riant

ai

03

a;

03
ai, 02, as,

equated to zero
4. If a,
ft,

is
;

a function of the differences of

and

'

a',

be the roots of the equations

d=
V"

0,

afx*

+ IV x +

express in terms of the coefficients the function

OB

- 7) 2

(a

~
') (

- P) + (7by
0,

2
)

(0

- a') (3 - ft) +

(a

- 0)

(7

- a')

(7

- /3')-

Denoting

this function

we

easily find
c
2
)

b'(ad

- be) +

c (ac

- b*) }

is

Attending to the definition at the close of Art. 150 we observe that this function an invariant of the two equations for it involves all the roots of the cubic in the
;

second degree, and all the roots of the quadratic in the first degree. If, in fact, we make the substitutions of Art. 150, and render the function integral by multiply2 ing by IT U", the result will not contain x, and is therefore an invariant of the
system.

The geometrical interpretation of the equation <f> = is that the quadratic V" should form with the Hessian of the cubic U a harmonic system.
u.

If a,

7;

a',

',

7'

be the roots of the equations

d =

0,

and prove that

express the following function (when multiplied by aa') in terms of the coefficients, it is an invariant of the system
:

344

Covariants and Invariants.

or, differently arranged,

(a-a')(ft-y')(y-ft')

+ (a-ft')(ft-a')(y-y')+(a-y')(ft-ft')(y-a) Am. 3 { (ad' - a'd) - 3 (be' - b'c) }

y', 8'

6.

If

o, 0, 7, S

',

ft',

be the roots of the biquadratics

0,
(a',
b', c', d', e') (x,

(a, b, c, d, e) (*, 1)*

=
(5

1)*

prove
aa'~2 (a

- a) (ft - ft')

(y

- y')

- 5') = 24 {ae' + a'e - 4 (bd" +

6'd)

+ 6cS},

and show that

7.

this function is

an invariant of the system.

its

Prove that the following function of the roots of a biquadratic and quadratic
geometrical interpretation
:

gives an'in variant of the system, and determine

1
ft

+y +
5

fty

a
a'

a5
aft'

+ ft'

Formation of Covariants by
where
viz.,

the

Operator D.

345
. . .

is

the result of

making x =
fl

in

F( Un

Un^

Z7

),

F
and
-D

= F(a ny
rf

_i,

...

flrj,

=#

rf

-+A 2#i

rf

tf

+ 3a 2

-+...
process, the source

In forming a covariant by this which we set out is altered by the

with

successive operations
, . .
.

till

we

arrive at the original function

F(a Q a^

#), from which

the source was formed. Since this

is

a function of the differences,

the coefficient derived by the next operation vanishes, and the covariant is completely formed. The corresponding operations S

on the symmetric function

degree in the roots

i//

have the

effect of

reducing the

step, the final symmetric function containing the differences only. Thus the successive a number of between the roots and coefficients operations supply

by one each

relations equal to the

number

of coefficients in the covariant.

is plainly equal to the number of times & operates in reducing i// to 0, i. e. equal to the difference And since of the weights of the extreme coefficients.

The degree

of the covariant

l
,
i

1
...

-,
a2

a
. .
.

is n-& - K, where K is the the weight of a n) ; weight of (ai, a 2 hence the degree of the covariant whose leading coefficient is
i/i ,

a<?<t>

is n-&

2c, the

same value as before obtained.

We

add two

simple examples in illustration of this method.

EXAMPLES.
1
.

Form

the Hessian of the cuhic

a a?

+ 3ia;2
Q a->

-f

Za z x + a z

0.

Taking the function

H=a
fl

2
i
,

we

find, as in Art. 151,

2o

08

- 7) 2 =
5,

18 (aj

ai a 3 ).

Operating on the left-hand side by

and on the right-hand

2

side

by D, we obtain

22a(/3-7) =

346

Co variants and Invariants.

and operating in the same way again,

o
2

22

(j8

y)

36 (i 2

Z)

The next operation causes both sides of the equation to vanish. quired covariant is, as in Art. 151,

Hence the

re-

"We find
roots.
2.

at the

same time the corresponding expression in terms of x and the

Form

the Hessian of the biquadratic

The

covariant whose leading coefficient

is

H= a
=
2,

ao- 0i 2

is called

the Hessian
2/c

of the biquadratic.

Its degree is 4, since tz

and K

and

.'.

ntt

4.

Changing the coefficients into 2 and we easily find 04 02 - 3

,

their complementaries, the source of the covariant is

2
i )

#4

-f

2(

<? 3

a<2,

+ 2 (a\ ai
154.

s)

x+

(#2

2
)

Theorem.* In the discussion of covariants through the medium of the roots, as in the previous Articles, the following
proposition, due to Mr. Michael Eoberts,
is

of importance

Any function
Let

of the differences of the roots of two covariants

is

a function of the differences of the roots of the original quantic.

(Co, <?

<72 ,

Cq )(x> y)

=C

(x

be two covariants of the quantic

Operating with

or 8 on the identical equation

Quarterly Journal of Mathematics, vol. v., p. 48.

Homographic Transformation applied

and remembering
/(flo

to

Covariants.

347

that, in general,

Df =

affy, where
a,,
. . .

ffi>

!>,

#) = ^^(ai,

a,,),

we have

and, therefore,
similarly

whence

proving that
di, 02, 03,

|3 r
a,,.

-7*

is

a function of the differences of the roots

...

155.

Homographic Transformation applied

Hitherto

to the

Theory of Covariants.
roots of equations.
different

we have

discussed the

theory of covariants and invariants through the

medium

of the

We
may

proceed

now

to give

and more general mode

of treatment,

some account of a by means of

which

be extended to quantics homogeneous in more than two variables, such as present themselves in the
this theory

numerous important geometrical applications of the theory. Although this enlarged view of the subject does not come within the scope of the present work, we think it desirable to show the
connection between the method of treatment

and the more general method referred to. give in the present Article two important
PROP.
I.

we have adopted With this object we

propositions.

Let any quantic TJn be transformed by the homo-

graph ic transformation

_ \x + = \~'~^~, x +n A
IJL

' '

if

I and

I' be corresponding invariants of the

tico

forms, we have

r-<y-A'i*v j.
To prove
this, let
tfo*

/=

S (ai

- a2 a )

(a 2

ST.

a,)

...

(<n

a*)',

each root entering in the degree

348

Covariants and Invariants.

Now, transforming
we have
p

the similar value of /', since x =

^ A

v~> x A

a <i

=
(X-\'ap )(X-\'a q y

Again, transforming
/ takes the

K9

and rendering the

result integral,

form

where
flo'

GO (A

A'ai) (A

- A'a 2 )

... (A

A'a*)

making

these substitutions for all the differences,

and

for of,

the denominators of the fractions which enter

by the

transfor-

mation disappear

and we have,

finally,

PROP.

II.

If $ (x)

be a covariant of the quantic

is

Un

the

new

value of $(x], after homographic transformation,

integral)

(token rendered

The proof

is

similar to that of the preceding Proposition.

We

have

this expression being obtained

by

substituting
a n for
ai,

#-ai,

QI,

...

a2

an

in the source of the covariant $(x) expressed in terms of the roots. Now, transforming, as in the previous Proposition, the value of 0(a?) thus derived; since the factors A X'ai, A - A'a 2
,
. . .

the same degree w in the denominator (for each root enters the source in the degree w), they will all be removed by
all enter in

the multiplier

</%

and the transformed value

of

(x) is

Reduction of Homographic Transformation.

349

156. Reduction of Homographic Transformation to a Double Linear Transformation. With a view to this

reduction let the quantic be written under the homogeneous

form

Un = ax n + na^y + "^
.

a*

xn -*yz

+...+an yn

and, in place of putting as before x =

fractions to

X +

^, x + A
y

-fJL

and removing
,
,'

/i

make

Un

integral, let

now - =

Xx+py
The

where y
transfor-

and - are the

y mation

variables in the ordinary sense.

may

therefore be reduced to a linear transformation of

y,

both the variables x and

the original quantic

and can be

effected

by putting in

x=

\x+ny,
from
a

y = \'x'+

p'y',

the introduction of fractions being in this

way

avoided.
of

Thus we

pass

homographic

transformation

functions of a single variable to the linear transformation of homogeneous functions of two variables.

The determinant
ficients

\/LL
\'IJL,

whose constituents are the

is

coef-

which enter into the transformation,

are

called the modulus

of transformation.

We

now

enabled to restate Propositions

I.

and

II. of

Art. 155, in the following

way
is

PROP.

I.

An

invariant

a function of the
is

coefficients

of a

quantic y such that when the quantic

transformed by linear trans-

formation of the variables, the same function of the new coefficients is equal to the original function multiplied by a power of the modulus
of transformation.

PROP. II.
quantic,

co-variant is

a function of the

coefficients

of a

of transformed by linear transformation, the same function of the new

variables

and

also

the variables, such that

when

the quantic is

and coefficients is equal to the original function multiplied a by power of the modulus of transformation.

350
The

Covariants and Invariants.

definitions contained in the preceding propositions are

plainly applicable to quantics homogeneous in any number of variables, and form the basis of the more extended theory of
co variants

and invariants referred

to in the preceding Article.

We

give

among

the following examples an application in the

case of a quantic involving three variables.

EXAMPLES.
1
.

Performing the linear transformation

x
if

= \X+/j.Y,
cy*

y=
CY*,

ax*

+ 2bxy +

= AX.* + 2BXY +
\ip)*(ac
if

prove that

AC-B* =

(\fjLi

b*}.

2.

Performing the same transformation,

(a, b, c, d, e)(x t

yY=(A,
2

B,

C,
4

D,
(ae

prove that

AE - BD + 3C
+ Ibxy +

(\/*i

Ai,u)
if

3.

Performing the same transformation,

ax*
cy*

= AX* + 1BXY + CT 2 = A\ X* +
(A^ui

and
a^x*

2bixy

ciy*

2Si

XY+
+
a\c

Ci Y*,

prove that

ACi

-f

AiC- IBBi =
1,

\ifj.}*(aci

2bbi).

This follows from Ex.

applied to the quadratic forms

by comparing the

coefficients of K

on both

sides.

Whence we may infer that, if two quadratics determine a harmonic system, the new quadratics obtained by linear transformation also form an harmonic system. For their roots being a, and 01, /h, we have
aai
4.
{

(a

- ai)03 - 0i) +

(a

- 0i) (j8 -

01)}

If the

homogeneous quadratic function of thrqe variables

ax*

by*

cz*

2/yz

+ Igzx + 2hxy

be transformed into

AX* + BY* +
by the linear substitution

CZ* + 2FYZ+'

1GZX + 2HXY

Linear Transformation.
prove the relation

351

H
B

H
G

F
C

352

Covariants

and Invariants.
;

fining our attention to the terms which multiply h

also that

dA -all

observing

= ra r-\ when terms are omitted which would be

2 3 by A A
,
,

multiplied in the result

&c.,

we have

which must hold whatever value h

d<b

may

have

hence

y-^ 9
dx

d$ ! -~-- + 2a -: = a d<j) + 3a 2 dQ -j- + da z da 3 da^

l

...

d$ + na^i -fda

(1)

and, substituting for $ the value

(B

19

.B2 ,

...Bm )(v,y) m

we have
-1
!/

m(m-

DB
=

xm

mD
we have the
lf
. .

whence, comparing

coefficients,

following equations

DB
the source

0,

DB, = J5

DB, = 2B

DBm = mBm^
coefficients

which determine the law of derivation of the

from

Bm

the leading coefficient

being a function of

the differences, since

DBo

0.
is facilitated

The
ciples
:

calculation of the coefficients

by the

follow-

ing theorem which has been proved already on different princoefficients of a comriant equally removed from the extremes become equal (plus or minus] when in either of them a a^ an are replaced by a ny a n_^ a^ respectively.
, . . . . .
.

Two

To prove
substitution

this, let

the quantic be transformed

by the
1.

linear

x =

OX +
4

Y,

y =

X+
n

F,

whose modulus = -

Thus
(o,
fli,

.,

)(*, y}

(a m a n_

a^2

aQ )(X,

Y)

n
,

and, by definition, any covariant

Z);

Linear Transformation.

353

whence it follows that the coefficients of the covariant equally removed from the extremes are similar in form, and become
identical (except in sign
their

when

K is odd)

when

for the suffixes

complementary values are substituted.

similarly that a covariant satisfies the diffe-

We may infer
rential equation
dd>

x T-

d<j>

ofy

dd>

dy

da n .i

da n.z

a,^

aQ

as well as the equation (1) already given. a n ] be an invariant of the quantic, Again, if (j>(a a ly a^ of transformation the present Article gives, employformer the
,
. . .

ing the definition of Art. 156,

tf>

0i> ^2,

... a n )

(A

l9

Az, ...

An

and proceeding as before in the case of a covariant, we prove that an invariant must satisfy both the differential equations
a^-r- + 2
ddi
dd> dd>
I

dd>

dd>

-jda-2.

+ 3^2

-fdas

+ ... +

nan-i

-~
da n
dd>

0,

dd>

an -jZ- + 2a^i
either of

d<j> yZ. + 30n_ -^ +

2

dd>

nth

~=
X

0,

if

may be regarded as contained in the other, since the linear transformation x = Y, y = (whose modulus = - 1), we have from the definition of an invariant
which

we make

proving that an invariant is a function of the coefficients of a quantic which does not alter (except in sign if the weight be odd)

when

the coefficients are written in direct or reverse order. the nature of Covariants and Inva-

Having now explained

riants of quantics, and the connexion between the two modes in which these functions may be discussed, we proceed to prove certain propositions which are of wide application in the formation of the Covariants and Invariants of quantics transformed

by a

linear substitution.

The student who

is

reading this sub-

2A

354

Covariants

and Invariants.

ject for the first time

may pass at once to the next chapter, where the principles already explained are applied to the cases of the
and
quartic.

quadratic, cubic,

158. PROP.

I.

Let any homogeneous quantic of the n th degree

by the linear transformation

f(x

y) become

F(X, Y)

also let

any function u of x, y become

Uby

the

same transformation ;

then we have

f (du _d\_

IdU

dU\

where

is

the

modulus of transformation.

To

prove this proposition, solving the equations

we have

MX = fix - M, MY= - \'x

whence

^ M dx
Again,
dx

=fjL,

M M
dy

=-n,

nr v M-=-\, dx

dy

du_dUdX ~ ~ y + dUdY_lf,dU 1
dX~dx
~dY~dx

dX

du_dUdX +
dy~ dXdy
which equations

dIJ^dY__^(

dU

dY

dy~M\^
1

may

be put under the form

dU\

dU
~dx~*(M
X'JT

and

since

f(\X + M F,

+ f[Y) . F(X, Y),

Linear Transformation.

355

changing

and
is

dU

into

jf

an <l

~rv-

dU
respectively,

the proposition

proved.

In an exactly similar manner, changing

X and F into

it

may

be proved that

The

results (1)

and

(2)

may

be applied to generate cova-

riants and invariants as we proceed to show. Suppose /(a?, y) and u to be co variants of any third quantic t', where may become identical with either as a particular case also, denoting by FC (X, Y) and Uc the same 'covariants expressed in terms of the X, Y variables and the new coefficients
-v
;

of v after linear transformation,

we

have,

by Prop.

II., Art. 156,

the identical equations

M*F(X, Y)

F (X,
C

F), and

MU^U
(1),
c

whence, substituting from these equations in

du

du\_

(dUc

dU \

-foJproving that /(
^

(dY' ~~dl)'
v.

du\
I

is

a covanant of

dx

And

in a similar

manner

it is

proved from

(2) that

leads to an invariant or covariant of

#,

according as u

is

of the

n th or any higher order.

add some applications of riants and covariants.

We

this

method

of forming inva-

2x2

356

Covariants and Invariants.

EXAMPLES.
1

If

be substituted for x and y in the quartic

itself,

(a, b, c, d, e) (x, y)

If,

and the resulting operation performed on the quartic

show that the invariant

I is

obtained.

We find
(,
b, c, d, e)

-~

V=

48

(00

4M +
,

3*2).

2.

(see

Ex.

x Prove, by performing the same operation on 2, Art. 153), that the invariant /is obtained.

the Hessian of the quartic

Here we

find

( a,

6, c, d, e}

(i, - -^

VH

72

(at*

2bcd

ad*

eb*

c).

3.

Prove that

(a, b, c,

d)

-,

-\G = x
(a, b, c,

where
4.

Gx

is

the cubic covariant of the cubic

d}

(x, y)

Find the value of

where w

(a, *,

<j,

rf)

(,

y)

3.

^*w.

9JT*.

159. PROP. II.

the

If 0(^
.
.

>

^i, ^2>
TO ,

...

flf)

form

i,

2,

(, y)

^w^ w

<?^y

an invariant of quantic of the n or

tjl

any higher degree,

dn u

dn u

d nu

dn u

invariant or covariant of u.

To prove

this, let

and, transforming as in the last Proposition,

,

d
+ y

*'

d _ ~~*

d
+
b

d
'

Linear Transformation.
also,

357

transforming

we have

whence

(x>*-+Y'^U-(,'t
and writing
(A,
this equation
...

'!)";
the form
n
y'}
,

when expanded under

9

A, A,

D)(X'

T')" =

(</o,

rfi,

</,

dn )(af,

we

have, from the definition of an invariant,

>

,A,A,
(c?
,

.A) =
. . .

showing that

dh

d*,

dn ]

is

an invariant or covariant.

#', y' are transformed similarly, as in the ar, present Proposition, they are said to be cogredicnt variables.

When

y and

EXAMPLES.
1.

Let the quadratic oo*2 + 2ai*y

oay

become

A X* + 2AiXY+ A T 2
2

We have then,
Now
since

as in Ex. 1, Art. 156,

it

follows from the last result, considering X',

Y and

x',

y as
/

variables, that
\ 2
j

d*Ud*U
This covariant
2. is called

d*U

\ z

(d^ud^u
[*r d^

"

d*u

\*5^/

)"

the Hessian of U.

When

u has the values

(a,4,c,rf)(* >y )

and

(,*,,

4
rf,

)(*. sO

what covarianta are derived by the process of the

last

example

^*.

(1).

(ac

I-}

x*

(ad

bc)xy

(bd

c*)y*.

(2).

(ac

358

Covariants and Invariants.

160. PROP. III.

If any

invariant of the quantic in #,

be formed, the coefficients of the different powers of k, regarded as

homogeneous functions of the variables

For, transforming
(0
,

x',

?/',

are covariants of U.

0i,

2,

an )
y'

U by linear transformation, let n = A n (X, (x,y} (A* A A

ly

Z)

F)

n
;

also if x,

y and

#',

be cogredient variables,

xy'-x'y =

Whence
(tf
,

M(XY'-X'Y).
w
(#, 2/)

0i, 02,

0)

*(a?y

w
aty)

becomes when transformed

(A,,

19

A,,...

An

n (X, Y) +

kM n (XT' - XT)-;
we have
(1,
>

and forming any invariant = (0, *i, fr, &>)(!,

.

of both these forms,

JF (*, *

x,

2,

*,)

Jf"*),

proving that
or that
<p r

is

a covariant.

When
manner
If
<j>
:

(xy'

xy]

is

replaced

by

(&o, Ji,

^> 2

bn ) (x, y}

n
,

we

have the following Proposition which

(a
2,

is

established in a similar
M
(a?,

i,

...

n ) fo

0w invariant of(a

2,

n)

y)

0// ^Ae coefficients

<j>

of k in
(0

^^o, 0i

#&i,

an + kb n )

are invariants of the system of quantics

(0
or,
,

0i,

2,

... 0n)

n
(a?,

^)

(fto,

*i, ^2,

... fc)

n
(a?,

y)

which

is

the

same

thing,

are invariants of the system.

If, further,

^ be replaced by a covariant, we

may

in like

manner generate new

this case.

covariants, a similar proof applying in These results hold for any number of variables.

Linear Transformation.
161.
quanticSj

359

PROP. IY.

//

0(a?,

y}

and

i//

(x t y) are homogeneous

the determinant

dx

dy

dx
i*

dy
the linear substitution

a covariant of these quantics.

For, transforming $ and

\p

by

x =

\X

we have

gvng

dx

Whence

dX dY

+ X^ du dx A'^,

&+"?
dx

du

dX dY
which reduces to
fd

-r-

+A

dx

-r-,

+
dx

fj.

dy

dy

d_<ty

dj>\
1

\d* dy

dy dx)

and the proposition

This covariant

is

proved.

is

called the Jacobian of

and

i/>,

and

is

often

written under the form e/(0, ^/). The Jacobian of n functions in n variables is a determinant of similar form, and can be

shown

to be a covariant

We

now

to illustrate

by an exactly similar proof. conclude this Chapter with some examples selected the foregoing theory. The student is referred for

further information on this subject to Salmon's Lessons Introductory to the Modern Higher Alyebra, and to Clebsch's Theorie

Der Bindren

Algebraiachen For men.

360

Covariants and Invariants.

MISCELLANEOUS EXAMPLES.
1.

From

the definitions, Art. 150, prove that

all

the invariants of the quantic

:

U(xy'

x'y) are covariants of U, the variable being x'

y'.

derive the covariants of a cubic from the invariants of a quartic expressed in terms of the roots.
2.
'

Hence
If

7i,

Iz,

/s,

... In
,
. . .

be the same invariant for each of the quantics

x
an

ai

x
<(>(x)

az

x
0,

of the order CT, where

en, az.

On are the

03

roots of

prove that

is

a covariant of $(x). 3. If ... on cii, 02, 03,

(a

be the roots of the equation

,

ai,

az , ... a n )
</>m

(x, 1)

=
2,

and

if

00w tf>l<J>2

JF(0,

1,

On),

where

0i, ^2,

0m

are all the values of a rational and integral function of

some

or all the roots obtained

by

substitution, find the equation

0.

whose

roots are the

(h

values

oi-^-, given
o,

82 <J>

Am. F(U ,U
4.

lt

Z7 2 , ...

UH =
)

0.

Denoting by

)8,

y,

and

a', )8',

7' the roots of the cubic equations

1)3=0,

prove that the following covariant of the system

expressed in terms of the coefficients

is

18
5.

- 2bb') x* + { (ac' + as

(ad'

+ a'd - be' - b'c] x + (bd' + b'd- Ice'} }

Express the identical relation connecting three quadratics in terms of their

invariants.

Let

U = aixz +

z 2bixy + ciy

we have

362
9.

Covariants and Invariants.

Prove that the functions
o02
0i
2
,

0004

4ai3 + 3022

0o 0s

which depend on the

differences of the roots of the equation

(o,

01,

2,

...

)(*, l)

=0,

give rise to covariants of the degrees

In -4,
10.

2n-

8,

6.

penultimate term in the equation of the squares of the differences of any quantic leads to a covariant of that quantic of the fourth degree in the variahles.

Prove that the

coefficient of the

11.

are

<f>

and

Prove that the product of two covariants of the same quantic whose sources he written under the form ty may
.D 2

^) + &c ---Mr. M. Roberts.

12.

Prove that the

m th power of the
(0o, 01, 02,

quantic

... 0n)

(a?,

!)'

may

be represented by
0n

+ *D (0n) +

!*(*.) + j-y-^ J(*,) + &C. Mr. M.

Roberts.

13.
is

Prove from both

definitions of a covariant that

any covariant of a covariant

be the roots of the equa... *)(*, 1)"

a covariant of the original quantic or quantics.

14. If
01, 02, 03, ...

a,
)(*,

and

j8i, jB 2 ,

3,

ft,

tions

U=

(o,

i,

3,

...

!)= 0,

and

Fs (J
U

*i, ^2,

from the simplest function of the

differences of their roots, viz.,

2(op

j8 8 ),

it

is-

and V. required to derive a covariant of the system This question will be solved if we express

-* (*-,)(*-*,)
in terms of the coefficients of

U and V.

For

this purpose

we have

and

if

17"

and

F"

be written as homogeneous functions of x and

y,

x-ay

dx

'

tfUog

U
'

x -i ay

dy

Whence, substituting these values

* (x

in the last equation,

we have
dy dx
'

ay) (x

- (3y)

_dx dy

which

is

the Jacobian of 17 and V.

It should be noticed also that the leading coef-

ficient of

J(V, V)

is

00*1

0i*o-

Miscellaneous Examples.
15.

363

To reduce
forms

the two cubica

(a, *, c,

Us
to the

d}

(*,

y),

V=
>

(', V,

c',

d') (*,

IdF
by means of a linear transformation

IdF

the coefficients in which are to be determined in terms of the coefficients of the

given cubics.

then

U= (a, r=

*, c,

d) (x, y)
ff)(x, y)
3

=
=

(A, B, C, D) (X, F)
(B, C,

3
,

(a', b', c',

D, E) (X, F)
,

3.

Now,
for

substituting the differential symbols

Dy
U,

- D x for

x, y,

and

,2)r M

M.

T^X

X and F in the Hessian of both forms of

we

find the operational equation

a
b

b
c

ABC BCD
we have
b'
c'

whence, operating on both forms of F,

a'
b'

d'

a
b

JY
b
c

bed
Similarly,

a
a'

b
b'

bed
JX
a
b' b'
c'

<f

b'

c'

d'

d'

where

/ is the ternary invariant of F. Again, since

V,)=I)r,
performing the operation
$>

and

-f(D w

-D r

(D y ,

Dr

$ (x,

y),

or

t//

(D v

- D,}

on equivalent forms we have

a
b
c

364

Covariants

and Invariants.

ficients of

"We are now in a position U and F.

to determine the coefficients of

F in terms of the coef-

For we have from former equations

V
Qx

c'

d'

Miscellaneous Examples.
The transformation
then identical.
If

365

plainly fails

when Q =

U+ icFhe a perfect
;

and Fbecome 0, for the values of cube the derived functions with regard to

x and y vanish simultaneously

whence we have the equations

~
b

b
c

Kb'

+ + KC'
to

KC

Equating these fractions separately

a
b
C

- K', we

find the equations

0,
0,

ttct

K 'b

+
+

KK'b'

+ +

K b' + K'C
KC

KKC'

+ K d + KK'd'=0;

and solving

for

*c,

K, KK,

we may

eliminate them, and thus find the condition that

U + K V be a perfect cube in the form

a
b
c

CHAPTER

XV.

COVARIANTS AND INVARIANTS OF THE QUADRATIC, CUBIC,

AND QUARTIC.
162.
riant,

The Quadratic.
if

The quadratic has only one invathe quadratic itself.

and no covariant other than

a and
]3

For,

be the roots of the quadratic equation

U 3 ax* + 2bx + c = 0,
riant or covariant are powers of

the only functions of their difference which can lead to an invaa- j3 of the type (a - /3) 2p ; the
coefficients in

odd powers of a - /3 not being expressible by the a rational form. Whence, expressing

H
by the
itself.

\a

-x

^ p-Xj
distinct

coefficients,

we conclude
b
z
,

that the quadratic has only the

one distinct invariant ac 163.

and no covariant
Covariants.

from

The Cubic and

its

In the present

Article the covariants of the cubic will be discussed as examples of the principles already explained, and in the following Article

the definite

number

of covariants

and invariants
is

will be deter-

mined.

In the

case of the cubic a covariant

obtained from a

function of the differences of the roots most simply

stituting
/3y

by sub-

ax,

ya +

fix,

a]3
;

+ yx for -

a,

j3,

y,

and

thus avoiding fractions

1
1

for, transforming a
(|3y
4-

|3,

we have

ax)

+ (ya +

The Cubic and

and when
fractions are

its"

Co variants.

367

removed we

arrive at the above trans-

formation (the order being equal to the weight in the case of or G) This mode of transeither function of the differences

forming functions of the differences will covariants of the cubic.

(1).

now be

applied to the

The Quadratic Covariant, or Hessian,

Hx

Transforming both sides of the equation

2

(a

o>|3

+ w 2 y) (a +
2
ct>

u>

/3

+ wy) = 9 (a? -

2 ),

we have
a<?{(a

+ w/3 +

y)#-f |3y

thus showing that

Lx +
are the factors of

^
z

and
(0 03

MX
-

(See Art. 59.)

H=
x

(0

a?) x

where
LI = j3y +
<*>ya

+ w 2 a]3,

MI = /3y + a> 2 70 + wa/3.

From
is

or from the relations of Art. 43,

the form of the Hessian in terms of the roots in Art. 151, we conclude that when a cubic
coefficients

a perfect cube, each of the

of the Hessian vanishes

identically.
(2).

The Cubic Covariant,

as in Art. 59,
2

Gx

We have,
3 fl

!(<*

+ w]3 +

o>

3 7) + (a +

w 2 /3 +

a>7)

= -27 (a<?a 3

-H

2^ 3 -30
we
find

2 J.

Transforming both sides of

this equation as before,

where
from

ferences

Gx denotes the covariant formed from the function of difG and operating as in Art. 153 on the source derived G (the sign being changed in order that G may be the
;

leading coefficient),

we

easily obtain

GX =
~

(0
2

#3

~ 3flf
- 303

102
2

+ 20! 8 ) tf3 + 3 + 20 3
3

(0 01 03 (0 3 030

01
2 2

- 20

3 ) 2

X9
X.

(0 3 #0

0l

-3

01

- 20 $ 0!

368

Covariants and Invariants of Quadratic,

,

fyc.

3 Resolving (Lx+ Ltf + (Mx +^fi) we

may
are

obtain the factors

are
/3

of

Gx

or,

more simply,
a +

since the factors of

+ 7 - 2a,

7+ a

1121121
2)3,

/3

27, the factors of

.

Gx
i

12
)3

-a?

-#'7 -a?

a-# fi-x'a-x

-x

y-x

when

fractions are removed.

obviously the following geometrical interpretation If three points A,B,C determined by of the equation Gx =
:

We have

= the equation be taken on a right line and three points with A\ B', C', such that A' is the harmonic conjugate of
;

regard to
of

and

1? of

B with
j5
;

regard to
0.

C and A

and C'
r

with regard to A and determined by the equation

then the points A',

&, C

are

Gx

(Compare Ex. 13,

p. 88,

and Ex., Art.

(3).

65.)

Expression of the Cubic as the difference of two cubes. This can be effected, by means of the factors of the Hessian,
as follows
:

(Lx + la)
For, as in Ex.
6, p.

(Mx + M^ = 27 U

114,

we have
()3

D-M* = v- 27
Transforming

y)(y -a) (a

|3).

this equation as before, the first side

becomes

(Lx + L,)*- (Mx

and

the second side

Substituting from previous equations,

we have

(4).

Relation between the Cubic and

its

Covariants.

The

following relation exists

A U\

Number of Co variants and

For, from Ex.
6, p.

Invariants of the Cubic.

369

114,
4J9r3 )=-27a 'A,

and transforming

this equation as before,

whence
(5).

A U* = Gx +
z

Solution of the Cubic.

The

expression

(J7v/A +
is

ft)*

(CVA
and

ft)*

a linear factor of

7".

For, from the relations in

(2)

(3),

we have
ft),
;

2a *(Lx +

L>Y= 27(Z7\/I s

2tf

(Ifo +

.a/i)

= 27 ( V v/A + ft)

and

since

+
is

zo

(jfo

a factor of

the proposition follows.

is

This form of solution of the cubic

164.

due to Prof. Cayley.

Invariants and Invariants of the Cubic. The following method of determining the number of covariants and invariants of the cubic is similar to that employed by Professor Cayley for the same purpose
:

Number of

The

H and G
2

cubic has only two cavarianta, their leading terms being

;

and only one

2

invariant,

viz.,
3

the discriminant A, where

A=#

4J2"

3
,

or

&=a

d*

+ 4ac -6abcd+4:dP-3b z c*.

To prove this, let 0(a, ]3, 7) be any integral symmetric function of the differences of the roots (of order w), expressible by the coefficients in a rational form.

We have then

(Art. 36),

<ft(a 9

P,y)=F(a H,G)
9

(1)

(where r remains to be determined) ; and, in the first place, if $ can enter this equation in be an even function of the roots,

even powers only, since

is

an even function of the

roots.

370

Covariants and Invariants of Quadratic,

fyc.

Eliminating the even powers of

G by

means

of the relation

G* +

4F

= a2 A,

we show

therefore that in the case of an even function of the

roots equation (1) takes the

form

(fl>(a,P, 7

)=F(a,H

A),

which

may

be written

a-0 (, ft 7) = F. (a, #, A) +

S Ff(

* A)
,

(2)

where
It

OT is is

the order of

(a, j3, 7),

and

an integral function.

now

No function of H and A
For, suppose
vanish,

necessary to prove the following

exists

Lemma
by
a.
;

which

is divisible

Fp (H,

A) to be divisible by a

then making a

we have
*,(&',*') -Op 4c?6 3 - 36V, the values
is

where H' = - b\ A' =

a vanishes.

of

H and A when
;

This equation

plainly impossible

ing

by means

of the equation

for, eliminat-

=-

b ,c

and d remain in the

equation connecting

H' and
(2)

A'.

Wherefore equation
a*<j>(a 9

must assume the form

=

p, 7 )

(a,

H, A);

for the first side of the equation is expressible as an integral function of the coefficients ; therefore so must the second side
also,

and consequently the

fractional part disappears.

Now,

to extend this result to

first

odd functions

of the roots,

we

have only to multiply the

side of the equation

by

and the second

function, since

side

by 27 G,
even.

for

G must be

a factor of every odd

is

We are now in a position

as to the
is

to prove the original proposition

number
form

of invariants

and

covariants.

For

since a*

of the

according as $

is

an odd or even function of the

roots, it follows

Covariants of the Quartic.

in the
first

371

place that there cannot be an invariant of an odd the in roots, since (a, H, A) does not remain the degree same function when a, b, c, d are changed into rf, c, b, a,

OF

an even degree must be respectively ; and the only invariant of besides A, it a power of A, since if F(a, H, A) contained a or

could not remain the same function

when

the coefficients are

similarly interchanged. Again, the cubic has only two distinct covariants ; for it has been proved that every function of the differences 0?$ is of one
of the forms

F(a, H, A),

or

GF(a, H,

A).

considering these forms as the leading terms of covariants, every covariant must be expressible as

Now,

,,A),
that
is,

or

is expressible in a rational and integral and Gx along with U and A or in other words, there are only two distinct covariants. 165. The duartic. Its Covariants and Invariants.

every covariant

form in terms of

shown already that the quartic has two invariants, / and G of the diffeand J (see Art. 151). From the functions rences of the roots we can derive two covariants x and Gx and G for from the relation whose leading coefficients are

We have

we

derive,
2

by the

process of
2

Art 151,
3

S (a -

/3)

(x
2

y) (x
9

S)

= 48

(UU

W)
-

and, expanding

UU

- Vf

we have
fliOaX +
(a
4

(0

aflx* + 2 (0

+ 20!

+ 2 (!< - 0203 ) X

In a similar manner,

since
2

03

+ 2! 3 -

2u2

372
we

Covariants

and Invariants of Quadratic,

fyc.

obtain the covariant

which reduces to the sixth degree

follows
:

and
+

if it

be written as

Gx = A
we
find,

x& +

Aiz 5 + A 2 x* +

x*

Ax* + A

x+

6,

by expanding the

the source

6,

above, or more simply, by forming and performing the successive operations of

:

Art. 153, the following values of the coefficients

2fl 3
3

A =- a*a
5

A
A
2,

fluffs

Here

it

will be observed

that,

when

is
5,

determined,

AI, and

may

be obtained from
their

A^ A

and

by

and complementary the was of with what the in accordance whole, altering sign
changing the
suffixes into

values,

proved in Art. 152.

the following Articles to discuss the leading properties of these two covariants of the quartio. 166. anadratic Factors of the Sextic Covariant,*

We proceed in
As

Gx

the quadratic factors of

Gx

enter prominently into the

first place to find expresfollowing discussion, sions for those factors in terms of the roots of the quartic, and to

we proceed

in the

deduce their principal properties. Since the factors of G, expressed in terms of

a, ]3, 7, 8,

are

the factors of
~~

Gx
*'

are

obtained from these by substituting

for a > to

ju

^' u

j/

;~-fl'
jc/

*&

~>v~l? o 7
by a

&*

respectively,

and mul-

tiplying each factor

remove

fractions.

* See a Paper by Prof.

Ball, Quarterly Journal of Mathematics, vol. Tii. p. 368.

Quadratic Factors of the Sextic Covariant.

373

Whence, denoting these

rrf au = U(
l
:

factors
l

by
l

w, i\

w we have
y

V*-p
-

X-y
l

X-a
1

X1

ac = rrf

U
"

aw
which valuesjof

'

u,

r,

?r,

arranged in powers of

a?,

are

W=
r

(|3

7-a-

2(f3y

a%)x + j3y(a +

8)

a8(j3

+ 7),

(7

+ a

-/3-SX-

2( 7 a -|3S)a?+ 7(i3 + S)-j38(7

+o)i

(2)

w=a +3and, consequently, 32 Gx =

a*iivn\

From

equations (1)

we

easily find

and from

these

and similar equations we have

v fj.

\-fji

a
17, Art. 27)

where

A,

ju,

v have the usual

r

meaning (Ex.
\"\
2

and

consequently,

\2

f\

^2
-

A.

whence
- U v] U?

(w^\ -

/u

r -vA

v)

(w <y\

/u

r-vA - vj.

Since, as this equation shows, the factors

on the second side

are both perfect squares,

we may assume
if

-vA -

v =

f,

yu

- v

y\ - v = 2

374
we

Covariants

and Invariants of Quadratic,

fyc.

have, therefore,

w <v/A - /u =
v

Ui

+ u z ~,

\/v- fi
u, v, w, the quadratic factors

from which values we conclude that

of Gx
,

are mutually harmonic.

Expression of the Hessian by the Quadratic Factors of Gx Since

167.
.

combining the terms in pairs,'and noticing that

the quantities between brackets being u,

v,

w,

we have

which

is

the required expression for

x.

168.

Expression of the Uuartic

itself
(3)

by the Unaa symmetrical

dratic Factors of
value

Gx
/u,

From

equations

may

be obtained for

tions in [place of A,
Pi 9 Pz 9 PS of the
2

; for, substituting in those equav their values in terms of the roots

U
3

equation 4p

Ip +
a* (w*

J=
-

0,

we

find

(v*

tf)

= 16 (p,

p 3 ) U,

u*)

= 16 (p, - pj U,

from which equations, by means of the value

ceding Article,
(flti)

of

x in

the pre-

we
= 16

obtain
( Pl

U-

x ),

(avf

= 16

(p 2

U- Hf),

(4)

Expression of the Quartic,

tyc.

375

We

now make

the substitutions
v*

u z - A! X\

2
,

w* -

A3Z2

3 are the discriminants of w, v, w ; thus replacing three whose discriminants are t>, by quadratics X, F, each equal to unity. By means of this transformation the for ms of the quadratics are further fixed, and the identical rex,

where A
?r

A A
2,

w,

lation connecting their squares (see (1), Ex. 5, p. 361) is expressed in its simplest form. Calculating the discriminants, we
find

A, =

(j3

+y-

a- S) {/3y

(a

+
3

8)

7 a()3 +

S)

with similar values of

A 3 and A

whence we have
/

A = -(A1

u)(X-v),

A 2 = -Ou-v)Ou-A), A 3 =-( v -X)(vbecome

u).

Making

these substitutions, the preceding equations

(Pi

P) (Pi
/o 3 )

PS)

X~ = EX Z

Pi U,
(5)

(p z
(PS

(p 2

pi)

Y =HX - Pz U
^
2

PI) (pa

PS)

EX -

from which are

easily

x,

and

the identical equation connecting

deduced the following values of X, F, Z:

U and

U = PI X* + PZ Y* + P3 Z\
Zare

(6)

where, as has been proved, X, F,

three mutually harmonic

quadratics whose discriminants are reduced to unity in each case. The value of x may be expressed in terms of X, F, as folSince lows.

we

easily find

Gx =

376
169.

Covariants and Invariants of Quadratic,

fyc.

Resolution of the Qjiartic.

From

the equations

we

find

17= (Pl - p 2) Y*

have the values determined by equations (5) and breaking up these values of U into their factors, we have three ways of resolving U depending on the solution of the
,

where

X Y Z
z
z
,

equation

4p

Ip +

J=

0.

The
fessor

resolution of the quartic has been presented

by Pro-

Cayley in a symmetrical form which may be easily derived from the expressions already given for and x

For, since in general

2

l(aix
is

+ 2bixy + dy*} + m (a 2 x* + 2b 2 xy + c2 y2 ) + n(a*tf + 2b3 xy +

c3 y

2
)

a perfect square

when

S/ 2 (aid - b?) +

2mw (a

c3

3 c2

- 2b z b 3 = )
I
2

0,

IX + m Y+ nZ is a
X, Y,
reduced to unity.

perfect square

when

+ m* + n* =

0,

Z being mutually harmonic, and the discriminants of each

resolution of

The
/,

U is therefore reduced to

finding values of

n such that IX +

mY + nZ,
pi

or

m \PS - pi
j0 3 Z7",

being a perfect square,

to satisfy the

may

vanish when[

U vanishes
2

or in fact

two equations
-V//03

-pi + n v//i

=
/t>a

0,

n*

0.

These equations are plainly

/

satisfied if

m
v
/s

n
pi

The Invariants and Covariants of

whence,
finally,

KU\H

is

the square of a linear factor of the quartic U. If it be required to resolve the quartic ic7" -

\HZ

it

appears

in a similar

manner that

X vanishes being a perfect square, must vanish when K or, values of 7, m, n must be determined so as to satisfy the
equations
/
2

U \H

n*

0,

These equations are plainly

satisfied if

whence

+&>.is

the square of a linear factor of K

170.

U - \HX

The Invariants and Covariants of K U

the equations (6) of Art.
168,

\HX

Employing

and denoting

+ Z* by F,
to the

we may, by adding -

reduce

it

form R^

X + R Y' + R Z\
2
Z

V to \HX - icV, where R + R + R* =

l

When

this is done,

we
-

have the following reduced values of

~ (2/02/03

=
=

K(2pi
ic(2/> 2

p2
3

p3 + X
)

-p /QI

PI)
/o 2 )

= K (2/o 3 -

378

Covariants

and Invariants of

Quadratic,

fyc.

On

account of the similarity of the forms

and

BX
l

which are of a fixed type, we calculate the invariants and covariants of *Z7- \HX by simply changing p l9 p 2 p z into E^ R Zy R3 in the expressions for the invariants and covariants of U.
,

Therefore, since

and

E -R
Z

(p z

pi) ( K

ApO,

-R,=

(p,

we

find the following values for the invariants of K U-

\HX

If

we form

the covariants

H
-

(K ,

A ),
2

and

Gf( K> A) ,

of

4Q
(the

4/c

J/cX

+ J\*
K, A),

reducing cuhic rendered homogeneous in

we

find,

as

M. Hermite has remarked,

Again, to calculate the Hessian of

KU- \HX we reduce

,

by the

substitutions

P^X* +

^r

+ pfz* - -

i(iu+ jv)

- - \iu,

pSX* + p^Y* + pfZ* the

first

(IH*+ JV),

of

which follows from the equations

multiplying by piX pzY*, from the first by changing

,

/o 3

2
,

X\ F

2
,

respectively Z* into PI

and the second

2
,

p*Y p,Z\
,

Number of
In
K
this
:

Covariants and Invariants of Quartic.

find the following

379

we ' way

form

for the Hessian of

U-\Hr

which

may

be expressed in the form

Again, since

p - 27 j* = 16 fa - pj
and
transforming
/(*, A)

( P3

Pl

( Pl

Gx
p lt
/o 2 ,

= p3

^1* - 27 J XTZ-,
2
.

into
2

JR ly

R^
2

A,

we
(Kt

find

- 27J = ;, A)
(

Q (P - 27J

),

x)r

We
of k-Z7

have therefore expressed the invariants and covariants \HX in terms of the invariants and covariants of U.

171.

JVumber of Cot a riant* and Invariants of the

dual-tic.

We

proceed to prove the following proposition,

the iwo distinct invariants

which determines the number of these functions:

The quartic has only

I and

and G. two distinct covariants whose leading coefficients are This proposition asserts that every invariant is a rational and
integral function of

/,

and

/ and

integral function of U,

X1 following discussion is founded on principles similar to those already employed in the case of the cubic.
,

<7,

and every covariant a

/, J.

rational

and

Gx

The

that

Attending to the observations in Arts. 36, 37, it is plain if $ (a, /j, 7, ) be any integral function of the differences

of the roots expressible by the coefficients in a rational form, we have, in general, considering the equation with the second term

removed,

tri(a,p,y 9 $)-F(a,ff,I,0) 9
where

F is

a rational and integral function, and r remains to be

determined.

380

Covariants and Invariants of Quadratic,

if,

fyc.

And
find

in the

first

changing

their signs,

be an odd function of the place, and subtracting the two values of

roots
0,

we

2^0

(a,

ft 7>

8)

= F(a,

ff, 7,

0)

F(a y H,

/,-).

whence, eliminatplainly vanishes with ing the powers of G beyond the first by the identical equation of Art. 37, we have
;

This value of

a'0(a,ft 7,
It follows that every

$)

GF (a
l

H,

J, J).

odd function ^ of the

differences of

the roots

is

divisible

by

and removing C
1

this factor
side,

on the

first

side of the equation,

and

32

on the second

we have

where

fa is

an even function of the

roots,

and FI a

rational

and

integral function. proceed to prove, in the second place, if $ (a, |3, 7, S) be any even integral function of the differences of the roots, of

We

the order
that a?$

CT,

expressible
7, S)

by the

coefficients in

a rational form,

(a, ]3,

can be expressed as a rational and integral

function of a y

H,

/, J.
:

To prove
There

this,

a. For, a be divisible a. by Making vanish, we suppose F(H, /, J) f = - b\ 4bd + 3c 2 have F(H /', </') - 0, where f J = 2bcd - eb* - c* (the values of H, /, J, when a = 0) and as

exists

the following lemma is necessary no function ofJT, /, J which is divisible by

to

'

T=-

it is

between

impossible to eliminate b, c, d, e, so as to obtain a relation conclude that no relation such as J?', I ', J', we
;

F(H', /', J'} s exists and therefore there the form F(H /, J} which is divisible by a.
y

is

no function of

and proceed with the proof of the proposition since, as has been already proved in the case of an even function of the roots,
;

We

now

Number of
we

Covariants and In^irinnts of Quarts. 381

have, dividing

by
8)

r ~s
rt

%
ff, J,

ft y,

= J.(,

J) + S

F>( ff

'

>Jl.

Again, since the

rational
a,

first

side of this equation is expressible as a

divisible

and integral function of the coefficients not the second side must be a similar function of the
this,

by
;

coefficients

and

by the lemma just

terms as

^Ff (H,I, S p
-

established, is impossible unless such

J) '

disappear.

"Wherefore

and, finally, we have proved that pressed by the forms

s
rt

0(a,

)3,

7, S)

may

be ex-

GF(a,E,I,J),
according as
is odd or even. "We are now in a position
<j>

or

F(a, H,

I, J],

to prove the original proposi-

tion as to the

number

of invariants

and

covariants.

must disappear, since F(a, H, I, J) be an invariant, a and this function could not remain the same if they were present

For,

if

when
larly,

the coefficients are written in direct or reverse order. Simi-

no odd function such

variant.

It follows that every invariant is a function of

Again, the quartic have proved that every function of the differences a*$ of the forms

J) can give an in/and J. has only two distinct covariants for we

as

GF(a, H,

I,

is

of one

F(a,H,I,J)
Now,
riants, it

or

6F(a, H,

I, J).

considering these forms as the leading terms of covahas been proved that every covariant is expressible as

n I J)
9

or

x and Gr every covariant is expressible in terms of is this the and and with J"; V, /, proposition which along to be was required proved.

that

is,

382

Covariants

and Invariants of Quadratic,

6fc.

MISCELLANEOUS EXAMPLES.
1.

If

U be
j8i,

\ U+
2.

nGx has the

If 01,

any cubic, and Gx its cubic covariant, prove tbat the Hessian of same roots as the Hessian of ?7, A. and p. being constants.
yi be the roots of

Gx =

0,

prove that

d
where

tddd
71);
1.

<*>(>

ft,

i]

<t>i

(ai, 0i,

prove also that

801
3.
fies

= 8i =

871

=-

Prove that any covariant of a quantic, whose roots are

01, aa,

a,

satis-

the equation

~da.

-r,

dy

where 73
4.

is

the degree of

<f>

in the coefficients of the quantic,

coefficients that

and

*i

2a.

Find the condition in terms of the

two

cubics

U and V should

determine a system in involution, the roots of one cubic being the conjugates of the roots of the other.

In

this case the cubics

may

be written under the following form

V=dxz +
also,

SKCX*

3/c

bx

/c

writing the discriminant of

P *D

pU + Fin
Z

general in the form

+ p*M +

N+ pM' + D' = 0,

we

find in this case

whence the required condition

Given

is

DM' 2 - D'M* =

0.

(For this condition expressed in terms of the roots see Ex. 10, p. 344).
5.

U = (a,
find the relation

6,c,d)

(x, y)3,

V=

(a',

V,

c',

d') (x,

y),

which connects the

coefficients of these cubics

when

it is

possible

to determine the ratio

/*,

so that

In

this case

the Hessian of

\U+ ^V should be a perfect cube. \ff+ ^V must vanish identically; and writing
tfH; E= Lx~ + Mxy + Ny\

it

under the two forms

A2
where

H+

\fjiK* +

Kx = (ad +

a'e

- 2bb') x* +

(ad'

+ a'd - be -

b'c)

xy +

(bd'

+ b'd - Ice') yz

we have

Miscellaneous Examples.
and elimination A 2
following form
:

383
is

3
,

A/t, /t

from these equations, the condition

obtained in the

ac-b2
ae'+a'c-2bb'
a'c'-b' z
6.

ad -be
ad'

bd

<?

+ a'd-bc' - b'c
a'd'-b'c'

bd'+b'd-2cc'
b'd'

= Q=

0.

if*

Given two
;

cubics, f(x)

and

</>(ar),

which are equal)

prove that

Q=

when

the roots of f(x) being o, ft, 7 (no two of the roots are connected by the relation

+ (7Rationalizing,

(-

0.

we have

also, since

tion

comparing the coefficients of the different powers of A we can render the last equaan integral function of the roots, which again, expressed in terms of the coefficients,

takes the form

3 3 (3P} -27(P -27Q) =0,

or

Q=

0.

its

7. If a quantic have a square factor, prove that the same square factor enters Hessian.

8. If a quartic have a square factor, the covariant tuple factor. 9.

a,
ft,

Gx has

that factor as a quin-

If f(x)

and

<f>(x)

be two quartics with unequal roots, the roots of /(z) being

have

7, 5; prove that the condition that a quartic of the system \f(x) two square factors may be expressed as follows :

a-

\/<t>(d)

02

0.

82

10.

Determine the condition in terms of the

is

coefficients

that

the

quartic

have two square factors. n<j> (x) may \f(x) In this case the Hessian of \f(x) + H>(*)

equal to K (\f(x)
2
,

+
,

n<f>(x)},
;

from

2 thus A, KH which identity we have five equations to eliminate A A/t, /x both of coefficients in the equafourth the degree obtaining an invariant In of

tions.

384

Covariants

and Invariants of Quadratic,

fyc.

11. Prove that the resultant of

two quartics becomes a perfect square when the

9,

invariant I/& vanishes.

Rendering rational the determinant in Ex.

the squares of the differences of the roots,

and dividing by the product of

introducing the coefficients,

we

find,

IM = IK?
12.

64.R;

whence, &c., &c.

Prove that the sextic covariant

Gx

of the

quantic

<(>(x)

may be

written

under the form

13.

Applying the principles of Art. 171, determine the form of the sextic cova

riant of the quartic

\ U + p.Hx

14. Calculate the values of

H,

J,

G,

for the Hessian of a quartic.

15.

A function

</>

of the differences of the roots of the equation

(0,

1,

2i

n) (X, l)

arranged in powers of a n being

</>

Ap + pAp-iOn +

'.
J.

-dp.za nz

A Q an P
i,

prove that

D^- = - na n _ijAj^, and hence show that function of the differences so also is ^(^o -^i> -^2, .
5 .

if t//(o,
.

s,

r)

-<4r).

16. If the discriminant of a biquadratic be written under the

form

(A

Ah

A*,
is

.4 3 )

(4,

I)

prove that the discriminant of this cubic

where As
17.

is

the discriminant of (o, the equation whose

3
i)

2,

a) (iP,

I)

Form

roots are

where

01,

o2 ,

03,

... o n are the roots of f(x)

= 0,
-

the resultant

R of /(a;)

and <(#)

being given.

Change the

last coefficient b m of

^ (x)

into J w

p,

and substitute

this value for

b m in the equation

E = 0.

CHAPTER XVL
TRANSFORMATIONS.
SECTION
172.
I.

TSCHIRNHAUSEN'S TRANSFORMATION.

the general beading of this Chapter we purpose collecting several propositions which could not have been conveniently given elsewhere, and which are of importance in

Under

connexion with the subjects discussed in the foregoing pages. commence with a general theorem relating to rational

We

transformations.

Theorem.
tion

th of a root of an equation of the n degree can be reduced

integral

The most general rational algebraic transformato an the 1 at n most. transformation

rational

For every
f(x)

of degree function of a root a r of the equation

is

of the

form

where \ and

i/>

are integral functions

also

and the denominator */>( a O ^( a n)> being a symmetric function of the roots of /(#) = 0, can be expressed as a rational
function of the coefficients.
tegral form.

^W

Whence

^-~-( *W

is

reduced to an in-

Moreover, the numerator of the former fraction

metric function of the roots of the equation

is

a sym-

/M
x

ar of the coeffunction as a rational be expressed consequently ficients of that equation ; that is, in terms of a r and the coef-

0,

and may

ficients of f(x).

2c

386

Transformations.

Now, denoting by F(a r )

by
division

this integral

form of ^-^, we have

4>(a r )

where

(OT)

does not exceed the degree n - 1

which proves the

proposition.

In the particular cases of the quadratic and cubic it follows that the most general rational function of a root can be reduced to a linear function, and a quadratic function of that root, In the case of the cubic this quadratic function respectively.

may
lows

be reduced to another form which

:

is

often useful, as fol-

Denoting the quadratic function the cubic /(0) by <//(0), we have

by $ (0), and dividing

proving that

whence it appears that the most general transformation of a root of a cubic may be reduced to a homographic transformation. In connexion with the proposition here established it is easy
to justify the remarks made in Arts 59, 66, relative to the soluWith this tions of the cubic and the biquadratic equations.

and ^ be two rational functions of n quanobject in view, let tities 01, a 2 , . . . a n (which may be considered as the roots of an equation), each having only p values when the roots are interchanged in every way.
in order

Denoting these values of both functions

01,
02,

by
03,

.-.

0J>,

we

have, for every integer/,

0!

V + 02^'
it is

a symmetric function of the roots, since possible values which 0^ can take.

the

sum

of all the

Formation of
In
this
0!

the

Transformed Equation.

387

way we
+fr

obtain the system of equations

03

+ fa

=TM

01^1

+ 02^2

+ 03^3

where

I ,,

...

T^ are all symmetric functions of a,, a ,a

2
, .

3,

. .

a*.

Solving these equations, we find at once 0i expressed as a $p-i ; and therefore by the symmetric function of t// 2 ^ 3 ,
present proposition reducible to a rational and integral function of T//I of the degree p - 1, since i// has only p values considered as

a function of
referred to

ai,

a2,

an

Now
<//

(1),

when p =2, and n =

jt)

considering the special cases 3, it is proved that a

linear relation connects

of

and in terms of symmetric functions = 3, and n = 4, and when and are in a CM, (2), a connected to be similar manner shown by homographie
a2 a3
, ;

t//

relation.

173.

Formation of the Transformed Equation.

The

transformation explained in the preceding Article was first employed by Tschirnhausen for the reduction of the cubic and

proceed to explain the biquadratic. in general the equation whose roots are
(ai),

We

method

of forming

(a),

^ (a 3 ), .... ^ (a),

where ^ (x)

is

a rational and integral function of x of the degree

= a + atf

n-1.
Let
Eaising
(x)
(a?

to the different

powers

2, 3, ...

n in

succession,

and

reducing the exponents of a? in each case below n (by dividing by /(a?) and retaining the remainder), we have

.... 4

+ fa + Irf +

2c2

388

Transforma tions

Substituting for x in these equations each of the roots of the

= 0, and equation /(#) adding, we find, if Si, $2 S3 &c., denote the sums of the powers of the roots of the required equation,
, ,

Si

= na +

ciiSi

+ a z sz +.... + a w _iS n _i, +

2

$2 = nb +

bi8i

S2

+ .... +

bn.i8n.iy

Now, expressing Si /(#), we have S i9 S29

.

a,
.

sn _i

in terms of the coefficients of of the coeffito

S n determined in terms

cients of

(x)

and f(x]

we

are also enabled

by Art. 133

express the coefficients of the equation whose roots are 0(aJ, Sn and therefore finally 0(a) in terms of Si9 Sz , 0(az), in terms of the coefficients of and /(a?) ; thus theoretically (x)
. .

the transformation
174.

is

completed.

Second Method of forming the Transformed Equation. There is another way of finding the final equation in by elimination, which we now give. Since
be multiplied by x, #2 x n , and the exponents of x reduced below n by means of the equation f(x) = 0, we have in all n equations to eliminate dialytically the n - 1 quan~ xn tities x, x*, We thus obtain the transformed equation
if ~l

this equation

entering into For example, if f(x] = xn - 1, the diagonal constituents only. we obtain the transformed equation in the following form
:

in the

form of a determinant

of the n th order,

O/Q

d)

G/i

d>2

&n-l

0.

#1

#2

a3

-<f>

Although

these

methods of performing Tschirnhausen's

if

transformation appear simple, yet

they be applied to par-

Tschirnhausen^s Transformation applied

to Cubic.

389

ticular cases the result usually appears in a complicated form.

Professor Cayley, by choosing a form of the transformation suggested by M. Hermite, was enabled to take advantage of

the theory of covariants, and thus to complete the transformation for the cubic, quartic, and quintic. "We shall content ourselves with
results for the cubic

showing in an elementary way how Professor Cay ley's and quartic may be obtained.

175. Tschirnhauseii's

Transformation applied to the

Cubic.

Let the cubic equation

ax 3 + 3fo 2 + 3cx + d =

be written under the form

z3

3Hz

G=

and

let it

be transformed by the substitution

y = A +
If
Zi,

KZ

2
.

z2 , z 3

ing values of y,

be the roots of the cubic, and we have

2/2-2/8
2/ 3

yi,

y2

y*

the correspond-

= (*-*3)(ic-*i),
= (s3-2i)(c-z2 ),
(1)

-2/i

and consequently,
23/1

- y a - y3 =
?/!

(2*i

z2

z3) K

+ (2z 2 z 3 (2z 3 Zx

i
i

z 2 ),
(2)

2^/2-2/32^/3

(2s z

-S53-2JK+
2)

Z,Z 2

Z,3 3 ),
ZaZj).

2/1

- 2/2 =

(2s

- Si 3

(22!Z

- Z2 S3 2

Wherefore,

if

the equation in y with the second term removed be

we have from

equations

(1)

and

(2)

H'-H.,

G'=G

390
where

Transformations.

and

are the Hessian

K
3

and cubic covariant of

;

3HK + G

and the transformation

is

therefore completed, since y\ + yt + y*

can be easily determined.

176. Tschirnbausen's Transformation applied to the Uuartic. In this case we do not attempt to form directly the transformed quartic, but prove the following theorem, which shows how this transformation may be resolved into two others.

Theorem.

Tschirnhauseri* s transformation changes a quartic

U into one having the same

To prove
this, let
4

invariants as ITT +

mH

x,

and

therefore

in general reducible to the latter form by linear transformation.

the quartic

# + piX* + p 2 x* + p 3 x + jt? 4 =

be transformed by the substitution y = #o + aiX + az x* +

If

a^x*.
,

#4 be the roots of the quartic, and y^ y a a?i, #2, the corresponding values of y, we have

#3

y*>

y\

x\

From

these equations

we
.

proceed to show that

-!,

= Po + Qo (#2#3 +

#1#4),

where

and

involve the roots of the quartic symmetrically.

In the

first place,

we

find

where A has

its

usual value,

viz.,

xzx3 +

x^

and secondly,

since

we

find again

Reduction of Cubic

to

Binomial form.

391

Finally, since the other terms in the product are obviously of + QoA, we have proved that the same form as

whence
(ys

2/3)

(yi

yO =

(v

/i)

(P +

Now, introducing /oi, p 2 , p 3 , in place of X, /u, v, this and the and similar equations preserve their forms ; whence, altering Qo into similar quantities, we obtain the equations

(y
(y>

2/3)

(yi

- yj = 4
yO = 4 yO = 4

f/o 3

/oa)

(P (P

3 ),

yO

(y

(PI

PS)

(^ 3

- ^0 (P - Qp

which lead at once to the invariants of the transformed quartic and comparing their values with the invariants of K 27 - \HX
;

given in Art. 170, the theorem follows at once.

177. Reduction of the Cubic to a Binomial form by Tscbirnhausen's Transformation. Let the cubic
ax*

be reduced to the form y 3 -

V by the transformation
x<i -

y = % + PX +
If
a?!,
,

the transformed cubic,

#2 #3 be the roots of the given cubic, and y a root of we have the following equations to deterq
:

mine p and

= y ly

x? + px 2 + q =
x?

+px3 + q=

from which we

find

X*
X
l

392
Adding X + #2 + xz
L

Transformations.
to this value of p,

we have
;

p + Xi
it

+ x2 + #3 =

follows (see Ex. 25, p. 57) that there are only two ways of completing this transformation, as the values of p, q ultimately depend on the solution of the Hessian of the cubic.

17S. Reduction of the (tuartic to a Trinomial Form Tschirnhausen's Transformation. Let the quartic by
ax*

+ 4bx* + 6cxz + 4dx +

be reduced to the form y 4 + Py* + Q, in which the second and fourth terms are absent, by the transformation
z y = q + px + x
.

If

a?!,

#z #3
,

XL

distinct roots of

be the roots of the quartic also y if y2 two the transformed quartic, we have the follow;

ing equations to determine

p and

-f

+ q = - y l9
from which we find

x? +px* + q

And, adding

x^

+ xz + x3 +
+
#2

a?4

to this value of p,

we have

p+
hence,

Xi

#3

#4

by Ex

5, p.

130,

it

follows that there are three

ways of

reducing the quartic to the proposed form, the determination of which ultimately depends on the solution of the reducing cubic
of the quartic.

179.

Removal of the Second, Third, and Fourth

n'
h

Terms from an Equation of the

Degree.

We

begin

Removal of 2 nd 3 rd and
, ,

4<

Terms of an Equation. 393

which we
shall subse-

by proving the following

quently apply:

proposition,

A homogeneous function V of the

#1, #2, #3,
.
.
.

second degree in n quantities

xn can be expressed
this, let
:

in general as the

sum of n squares.
#1,

To prove
lowing form

F, arranged in powers of
2

take the fol-

F = Pi *1
linear

+ 2 &*! +
.

.#!,

where PI does not contain x^ x 2y

xn

also Qi

and

jR t are

and quadratic

functions,

respectively, of x2 , #3 ,

...#.

We

have then

also,

assuming

Fi
where

=1 ,

~ = P*rf
\

is

a constant, and

Q and
2

do not contain x and

l

we

have, similarly

so that

Proceeding in this way,

we

arrive ultimately at Rn-\

is

n-\

which

is

equal to

Now,

proved. returning to the original problem, let the equation be

;

Pn#n
l

and the proposition

+P&P-* +
3 -f

of

+p xn

~l

.+pn

and, putting

y = ax*
let the

fix

+ yX* + $X +

f,

transformed equation be
n n y + Q>y
-1

n- 2

0,

394
where, by Art. 173,
of the
first,

Transformations.
ft, ft,
. . .

ft,

are

homogeneous functions
e.

second,

r th degrees in

a, |3, 7, S,

Now,

if a, /3, 7, S, e

can be determined so that

ft = 0, ft = 0,
c

ft = 0,

the problem will be solved.

For
its

this purpose, eliminating

from

ft and

ft,

by

substituting

value derived from ft =

0,

we

obtain two homogeneous equations,

of the second

and third degrees

in a,

|3,
z

7, S

and by the pro-

position proved above

we may
u*

write

under the form

v*

+ w* - f,

which

by putting u = v and w = t. From these we find 7 = la + w/3, and S = Aa + wii/3 and simple equations these values in Q = 0, we have a cubic equation to substituting
is

satisfied

determine the ratio

quantities a,
)3,

)3

a.

Whence, giving any one

of the

7, S, c
is

a definite value, the rest are determined,

and the equation

reduced to the form

n n-' + y + Qty

Q&n

-5

ft,

0.

In a

similar

way we may remove

method

the coefficients ft,

2,

ft,

by solving an equation of the fourth degree.

Applying
this

to the quintic,

we may reduce

it

to

either of the trinomial forms*

x5 +

Px +

ft

x 5 + Px* +

or again, changing x into -, to either of the forms

x5 +

Px +
3

ft

x 5 + Px* + Q.
Serret (see his

In

this investigation

we have followed M.
I.,

Cours d'Algebre Superieure, Yol.

Art. 192).

* See Note A.

Reduction of Quintic

to three

Fifth Powers.

395

180. Reduction of the Qjiintic to the Sum of Three Fifth Powers. This reduction can be effected by the solution of an equation of the third degree, as we proceed to show. Let
(0
,

0i,

2,

03, 04,

where

/3i, j3 2 , /3 3

are the roots of the equation

p&
Now, comparing
00
02
4

j9 2 #*

+ ^^ + PQ =

0.

coefficients in the

two forms of the quint ic,

+
+
2 j3 2

#1

#2

&3
2

01

= =

&l/3l
3

#3/3 3 ,

= ^1/3^ + bfi* + ^/3 3

03
05

^/Sx

J 2 )3 2

+ kfr 3

whence
=
=
0, 0,

0.

When
equation

these equations are taken in conjunction with the

0,

we have

the following equation to determine

j3i, )3 2 ,

00
01
2

01
2

02
3

0.

04 05

03

04

Also,

bi,

Z>2,

are determined

by the equations

396
whence the question
known.
is

Transformations.
completely solved

when

j3i, /3 2 , /3 3

are

This important transformation of the quintic is a particular case of the following general theorem due to Dr. Sylvester
:

Any
reduced

homogeneous function of
to the form

x, y,

of the

degree 2n

1,

can be

th by the solution of an equation of the n degree. The proof of the general theorem is exactly similar to that

above given for the case of the quintic.

181. <luartic*s

Transformable into each other.

We

proceed to determine under what conditions two quartics can be transformed, the one into the other, by linear transformation.

Let the quartics be

U=(a,

b,

c,

d,

e)(x,yy = a(xy'Y ^ a'(x'a'y') (x'~

F= (',
and
if

b', cT, d', e') (*',

0V) (x'-y'y') (*'-

they become identical by the transformation

x = \x +

fjty,

f y = X'x +

//y,

we

have,

by Art.

38,

showing that the six anharmonic ratios determined by the must be the same for both equations.

roots

From

these equations

we have

also the following relations

:

between the invariants of the two forms

(1)

whence
i' 3
I

r
-,
(2)

Quartics Transformable into each other.

/3

397

The quantity absolutely unaltered by transformaJ being tion when the quartic is linearly transformed, is called the The condition expressed by absolute invariant of the quartic.
~,

equation (2) is, therefore, that the absolute invariant should be The condition here arrived at the same for both quartics.
agrees with the result of Ex. 6, p. 146, where it is proved that the sextic which determines the anharmonic ratios of the roots
involves the absolute invariant, and no other function of the
coefficients, of

the quartic.

The
illustrate

conditions expressed

always necessary ;
Suppose, in the

but not always

first

by the equations (1), (2), are sufficient, as we proceed to

by two exceptional
place,

cases.

where

u, v,

Although

are of the linear form Ix + my. 7 3 J'3 = -=,- is satisfied in this case, the the condition
r, ?/,
t?',

value of these fractions being 27, it is impossible to into F, since it is impossible to make vw a perfect transform

common

square by linear transformation.

Secondly,

if

U * u*v,

F ^w'

4
;

although the equations /' /' = 0, 7=0, J' = 0, J =

transform

= r4 /, J' = i*J are

satisfied,

since

0, it is, nevertheless,

impossible to

7 into F.
cases it

In both these
six

anharmonic
it

ratios

would be impossible to identify the depending on the roots of the quartics.

it is

In general,

may

be stated that

impossible to transform

one quantic into another by linear transformation when any relation exists between the invariants of one of them which
does not exist between the invariants of the other (see Clebsch's Theorie der Bindren Algebraischen Formen, Art. 92).

398

Transformations.

MISCELLANEOUS EXAMPLES.
1
.

Transform two given quadratics in

x,

to the forms

auz

bvz ,

where u and v are linear functions of x and

2.

y.

If the coefficients of three quadratics

be connected by the relation

02

cz

0;

03

<?

prove that they

may

be reduced by linear transformation to the forms

The determinant here written

is

the condition that the three quadratics should

determine a system of points or lines in involution.

3.

Reduce

(a,

b,

c,

d)

(x, y)

to the

sum

of

two cubes by the method

of

Art. 180.
4.

Prove that two cubics can in general be transformed one into the other by

linear transformation.
5.

Express three cubics,

Z7,

V,

W, by means
V

of three cubes.

Assuming

AZ7+/UF+
.and comparing coefficients,

W= (x-py)

3
,

(1)

we have

-f

vcs

These equations, by eliminating

values of x, p, y
:

in this

way we

A, /u, v, give three values of p, and corresponding obtain three equations of the form (1) to deter-

mine

F,

Win terms

of

Miscellaneous Examples.
It is easy to see that p is given

399

by the equation
ap
^zp
czp

03

=0.

dz

cap

d$

A
6.

similar

method may be applied

to express

th quantics of the n order in terms

of n n th powers.

Prove that the three roots of a cubic

may

be expressed as

where

and

From

Art. 60, putting

^ 2

/
ii

JT,

where

or

1,

we

derive

K(y - a) = ^ya + Hj. (y + a) + ^(a - j8) = Ha& + HI (a + )8) +

These homographic relations between the roots

-ff2 ,

(1)

#2.
be written in the form

may

where the numerator and denominator in

this being

6 are supposed to

be divided by 2Z"; and

done

it

will be

found that

/,

m,

lm'
3

I'm

=1 = 1 + m',

and the roots

a, 7,

m' are connected by the relations 2 may be represented as o, 6 (a), (a)

i',

being equal to a. It is important to observe that the equations (1) are conthat is to sistent, the sum of the expressions on the right-hand side being zero say, IT must have the same sign in all three, any other combination of signs being
(a)
;

inadmissible.
7.

ratios

Given a binary cubic U and x y and xf y'; prove that

:

(See Serret's Cours d'Algcbre Superieure, Vol. II., Art. 511.) its Hessian x , the cubic being satisfied by the

1 i

*T + <T
dx dy

V
is

xy'-x'y

an absolute constant,

Reduce

U to

the

sum

of

being the discriminant of U. two cubes by a linear transformation whose modulus

.

1,

and the constant may be easily shown to be

This

is

another form of

V-3

the homographic relation of Art. 60.

8.

nects

Prove that a rational homographic relation in terms of the coefficients conany two rational functions of the same root of a cubic equation but that
;

the relation

is

not rational

when

the roots are different.

400
9.

Transformations.
Transform the quartic
(a, *,

,*,)(*,!)*

into one

whose invariant

I shall

vanish.

Assuming

a;

-f

2^ +
2

and making the invariant / of the transformed equation vanish, we have

2 P3)

(4>-/>1)

=0,

(1)

where

known quadratic Expanding (1), we have

<J>

is

function of y, not involving

which determines

</>,

and consequently

77,

by means

of a quadratic equation

and

may have any value. By a similar transformation / can

reduced to the transformation

he made to vanish.

10. Prove that the most general rational transformation of a quartic f(x)

may

be

p-x

q-x'
show
that the second term

"When

P = Rf(p) /'(?),
is

and

Q = - Rf(q] f'(p),

of the transformed quartic

11.

absent.

Prove that the transformation

axz + l&x

+y
a homographic (3) a homo;

may he

resolved into the three successive transformations

;

(1)

transformation

a transformation of the roots into their squares graphic transformation.

(2)

12. If

he any integer, prove that

where 2o and Si are symmetric functions of

13. If
<f>(x,

x\ t #2, #3, #4.

y)

and

ty(x, y}

he two co variants of the binary form

of the degrees

.p

and

y,

respectively

and

if

,(x-!r, V
yrfy.

he expanded in the form

(Fo, Fi,

F2

....

F^KZ, T)

prove that

Fi, F2,

. .

Fp

are covariants of U.

Homogeneous Function of Second Degree,

fyc.

401

SECTION II.
182.

HERMITE'S THEOREM.

Homogeneous Function of Second Degree

have already shown, that a in a general way (Art. 179), homogeneous function of the second degree in the variables may be reduced to a sum of squares, no hypothesis being made as to the nature of the coefficients of the function considered.

expressed as

Sum

of Squares.

We

We

now

return to the con-

sideration of this

problem when

the coefficients of the function

are supposed to be all real ; and we proceed to determine, in magnitude and sign, the coefficients of the squares in the transformed function.

#n ) be a homogeneous function of the second degree in n variables with real coefficients and let us suppose that it is reduced by the method of Art. 179 to the form
9

Let F(sri #2

>

Pi

(a?i

where

all

the coefficients of this

new form

are real.

Making now

the linear substitution

l

Xi = x

we have

Since the modulus of this transformation is equal to 1, the discriminants of both these forms of .Fmust be absolutely equal.

402

Transformations.

Denoting, therefore, the discriminant of

F by
,

tt ,

we have

and

the variables ay+ i, #/+2 ... xn are vanish in both forms of F, we have
similarly,

when

made

to

Now, giving j
Pi = A!,

the values

1, 2, 3, &c.,

we

find

^3

2
,

Ai

^3

=,... ^n = A A^_i
-r
2

An

and the

coefficients are

determined in terms of the discriminant

of the original quadratic form in n variables and the discriminants of the forms in n - 1, n - 2, &c., variables derived from

the given form by causing one, two, &c., of the variables to vanish in succession in the manner just explained. %n) are Again, since the constants in the form F(x^ #2
,

than in a form composed of a sum of n linear functions of n variables, we learn that of can squares be reduced to a sum of squares in an infinity of ways. It is
in
1) less

number \n(n -

most important, however, to observe that

transformation
(affecting
is

in whatever

way

the

made, provided it is real, the number of coefficients these squares) which have a given sign is always the same.
is

This theorem, which

due to Sylvester,

is

easily proved

for

suppose the contrary possible, and let

where the number of positive

identity
is

coefficients
all

on both

sides of this

the terms positive, by those affected with transferring negative signs to the opposite sides of the identity, we shall have a sum of / squares identically

not the same.

Making

equal to a
/

sum

of

squares,

where

m is
,
. .

substituting such values

for

x l9 #2

Now, greater than /. xn that each of the

squares

may

ways),

we
is

vanish (which may be done in an infinity of find a sum of squares identically equal to zero,

which

impossible.

HermMs

Theorem.

403

183. Hermite's Theorem. The principles explained in the preceding Article have been applied by M. Hermite to the determination of the number of real roots of an equation f(x) = comprised within given limits. The special form of the function

F which he makes use of for this purpose is

r

n
2
a,-

2^
<*r

in

xn are any variables in number equal to the degree of the equation and r takes all values from 1 to n in-

which

Xi,

a?2 ,

clusive, the roots of the equation

being a lt a 2

a,t

also

is

any arbitrary parameter. This form is plainly a symmetric function of the roots of and as the coefficients of this equation the equation f(x) = are supposed to be real, F will be also real, when expressed in terms of these coefficients and /o, provided the parameter p be
;

a n are not all given any real value. If the roots m, a 3 a 3 will of not be form obtained the assumed real, by real trans, ,
.

formation

but

it is

easy to deduce from

of conjugate

it,

as follows, another

form which
If
ai

will be so obtained.

and a z be a pair
=
(cos a

imaginary
(cos a
2

roots,

we may

write
ai
>'

sin a),
l

a2

>*

. . .

sin a).

Denoting

for shortness

x + a r xz +

a,.

#3 +

+ a r n xn by
find

~l

rt

and

substituting these values in

and

2,

we

where

U andj F are real

-=
CL\

also putting
Oz

r (cos

$ +

sin 0),

= r

(cos

sin

p
ai

the part of the function

F depending on
-

and a

2,

viz.,

CLl~

2D2

404
becomes

Transformations.

which

may
\

be also written as the difference of the squares

|-

Fsin

|Y-

2;-f

CT S

in| + FoosgY;

proving that two imaginary conjugate roots introduce into two real squares, one of which has a positive and the other a
negative coefficient.

We now state Hermite's theorem

tion f(x)

(x

di)(#
:

- a2 )

.'(x

an)

and unequal

roots

if then by a

REAL

Let the equahave real coefficients substitution ice reduce

as follows
:

Y?
ai

ct 2

-p

a^-p
.
. .

an

p
*

(1)

n'here

x\

+ a r oc z + a rz %3 +

+ a rn xn

~l

^o

a sum of squares, the number of squares having positive coefficients

uill be equal to the

tion

= f(x)

0,

number of pairs of imaginary roots of the equaaugmented by the number of real roots greater than p.

This theorem follows at once from what has preceded if we consider separately the parts of the function (1) which refer to
real roots

and

to

imaginary

roots, for obviously there is a posi-

tive square for every root greater than p, and we have proved that every pair of conjugate imaginary roots leads to a positive

and negative

real square, without affecting the other squares

independent of these roots.

The number

of real roots

between any two numbers

pi

and

For, denoting in general by Pj ps may when p = pj, by Nj the the number of positive squares in number of roots of the equation f(x) = greater than /o/, and by

be readily estimated.

21 the number
whence

of imaginary roots,

we have

Pi =

JVi

/,

P, = N, + 7;

Ni - N, =

2,

Hermite 's Theorem.

405

proving that the number of real roots between p and /o 2 is equal to the difference between the number of positive squares when

p has the values pi and /o 2 respectively. The number here determined may be shown
,

to

depend on a
consider

very important series of functions connected with the given

In order to derive these functions equation. under the form [Art. 182)

we

The number
form which are

expresses the

number

of coefficients in this

positive, or, which is the same thing, the of the following quantities which are negative
:

number

A2

P
We proceed now to
,
.

(2)

calculate Ai, A 2 A n in terms A/, and as the method and the roots of the equation f(x] = is the same in every case it will be sufficient to calculate A 3 all the i. e. the discriminant of the original form of F when
. .

of p

variables except

a?i,

tf 2 ,

xz vanish.
ar

Writing

for shortness v r

we have

in this case

p
a,-

= 2v r (#i
of this

4-

a r #2 +
is

The discriminant

form

which

may
1
a2

be written as the product of the two arrays

1
CL\V\
2 V-2

a2

a2

406
and, consequently,
1

Transformations.

1
/ \a<i

\2

03)

^a 3

In an exactly similar manner we

find

^
where the notation

V (i>

02

a3

ay)

v (i, a 2 a 3 a/) is employed to represent the product of the squares of the differences of a it a 2 a 3 a/. A n are all determined. Hence the quantities Ai, A 2 A/
,
, . .
.

Now, multiplying

the numerator and denominator of each

of the fractions in the series (2)

by

p) , each value of

is

rendered integral, and the series becomes

F,

V
(p

F
(p
ai)

where

(/o

- a 2 ---)
(p

(/o

- a),
(/o

Fi =

- a2 )

- a3 )

----

- an ),

F2 =

SV

(ai,

a2 ) (p - a 3) ---a 2 , a 3 ) (p

(/o

- an ),
- an ),

2V

(01,

- a4 )

---- (p

2,

0.3,

Since negative terms in the series (3) correspond to variations of sign in the series
F",

Fi,

F F
2,

3,

Fn

it is

proved

that the

number

of variations lost in the series last written,

,

when p passes from the value /oi to the value p 2 is exactly equal to the number of real roots of the equation / (p) =0 comprised
between
^i

and p z

184. Sylvester's

Forms of Sturm's Functions.

It

will be observed that the functions F, Fi,

in the

2,

&c., arrived at

preceding Article have the same property as Sturm's

Sylvester's
functions

Forms of Sturm's Functions.

fact

407

by positive multipliers only, as was observed by Sylvester, who first published these forms in the Philosophical Magazine, December, 1839. The of the two series of be established functions as identity may
;

from which in

they

differ

follows

p.

"We make use of the notation already employed in Ex. 7, 312, and we propose to show that the Sturmian remainder
differs

RJ

From
where

only by the positive factor jj from the function P}. the example referred to, we have

Rj~Ajf(x)-Bjf(x),

(1)

AJ =

Ao + \\x
jUiiC

+ \ z z? +
+
ju 2

+
.

BJ = no +

2
-i.

and from the value

of r_,- there given

7J

we have immediately
and Vj
differ

*W =

SV

(eti,

a 2 , a 3 , ---- ay),

showing that the leading

coefficients in JRj

only

now proceed to prove that the last coby the factor jj. efficients in these functions differ only by the same factor.
For
in
it

We

this purpose, dividing the identity (1)

by /(a?),

substituting

from the equation

^M-^a-^-u "
/(*)

fL-i
X*

X*

and comparing

coefficients,

we

find

\%\ +

....

-f

A_i S_2

Also, putting x =

in (1),

we have

and, substituting for

/u

in terms of Ai, A 2 , A 3 , &c.,

~l~

- = AO

'S'_i

Ai'^o

408
whence, giving to A
calculation of rrt _y,
,

Transformations.
Xi,

....

Xy_i

the same values as in the

we

find

f-l

Now,
ting p

referring to the calculation of Ay in Art. 183,

0, or v r

and put-

=
ar
,

in the value of Ay there found,

we

find

for the determinant just written the value

^y(ai,
^rf

a 2 , a3 ,

ay)

aia-jOs

... ay

hence, giving

pn

its

value in terms of the roots,

a 2 , a3 ,
.
.

we have
. . .

r = (-

l^yy 2V (cu,

ay) ay+1 ay+2

an,

which was required to be proved. The first and last coefficients of Rj, when divided by yy, having been thus shown to be the same as in the form Vj it follows that all the intermediate terms must be similarly
related
;

for, in

the

first place,

Rj
.

is
.
.

a function of the diffe-

rences of the quantities #, 01, a 2 a n , as may be seen by before transforming f(x) calculating Rj by the substitution
2
is

= a x +

#!,

as in

Ex.

3,

Art. 92.

When

this transformation
9

completed, every 7?y, of the differences consequently, Rj

;

coefficient in

as well as z

is

a function
differential

satisfies the

equation

dai

+ -^ +
da 2

n A -7- LB/ = 0, dan j

d\

dJRj or-r^-

dx

0,

showing, as in Articles 138 and 157, that all the coefficients may be obtained from the last by a definite law. The same
conclusions plainly holding also for the function P}, fore proved, finally, that
it is

there-

Sylvester's

Forms of Sturm's Functions.

409

EXAMPLES.
1. To reduce two quadrics in three variables to the sums of the same three squares with proper coefficients.

Let

U=axz + by* + cz z +

2fyz

2yzx + 2hxy,

"We have then identically

\h+hi
1

X
T
AC +

MA")

V + /i
X
T

CI

is the discriminant of \U + V; and *(A) is a function of the 2nd degree in A, the symbols X, Y, Z being retained in it for the present, and not replaced by the values involving A.

where A(\)

Resolving into partial fractions,

we have
*(A 2 )

*(A 3

A'(Ai)

A-Ai
(As)

A'(A 2 )

A-A 2

A' (As)

- AS'

(1)

in

which *

(AI),

(\z),

are all perfect squares, since they are obtained

by
re-

bordering the vanishing determinants

(AI),

A(AZ),

(Aa).

(See Art. 129.)

Now, replacing X, T, ducible to the form

by

their values,

\U\ +

FI, &c.,

(Ay)

is

easily

a\

+ g\

-(A -A,)*

where j

1, 2,

or

3,

and

Uj is

independent of A.
(1),

Substituting these values in

we

find

-A

A
(A 2 )

-r

(Ai)

(A 3 )

410
Equating the

Transform a tions
coefficients of A,

we have

A'(A 2

~~

A A

>/>.

'

(Ai)

A'/-k A (A 2

'

>

A '/I \' A (As)

which was required

to be done.

It is to be observed that this

reduction here given is due to Darboux the number of variables.

2.

problem has only one solution. The mode of and is plainly applicable whatever be
;

transformation to a

Prove that a quadric in n variables sum of n squares.

may

be reduced by a real orthogonal

An

modulus written

orthogonal transformation is a linear transformation such that, when the as a determinant is squared the terms in the principal diagonal
of the squares of th&

are each equal to

1, and all the other terms vanish. In a transformation of this kind it follows that the sum

new

variables is equal to the

sum

of the squares of the old.

3.

"Writing as before one of Sturm's remainders in the form

prove that
Sj-l

Sj-Z

where
4.

s xJ-

gj.\.

Denoting by UH

S (* - a
r=l

r)

(*\

+ axz + a2 3:3 +

prove that the discriminant of Uj

may

be determined by the equation

where Aj and

yj

have the same

signification as before

Aj =
value of

for a certain value of x, Aj.\ and Aj+\ have opposite signs for the
x.

and show directly that if same m in the

NOTE.

Hermite's theorem holds where a r

is

changed into

(a r

p)

enunciation on p. 404,

being any odd integer, positive or negative.

Transformation of Binary

to

Ternary Forms.

411

SECTION III.
185.

GEOMETRICAL TRANSFORMATIONS.*

Transformation of Binary to Ternary Forms.

think

it desirable, before closing the present Chapter, to a brief account a of transformation from a binary to give simple a ternary system of variables, whereby a geometrical interpre-

We

be given to several of the results contained in the preceding Chapters. The applications which follow in connexion with the quadratic and quartic will be sufficient to explain this
tation

may

mode

of transformation

and

will enable the student acquainted

with the principles of analytic geometry to trace further the analogy which exists between the two systems.

Denoting the original variables, nary system, by ^ ?/ we propose

,

i.e.

the variables of the bi-

to

transform to a ternary

system by the substitutions

x = #

2
,

y = 2# y

= y*.

For example, taking the simple

roots are
a, )3, viz.,

case of a quadratic

whose

(a

J3);r

y + a(3y* =

0,

and transforming, we obtain

*-*(

+ j3)y +

a/3s

0.

(1)

We
This

have

also the identical equation

y - 4zx =
1

0.

is the equation of a conic, which we call V, and (1) is the plainly equation of a chord of this conic joining the points a and /3, the point determined by the equations

x 2

= y
<37=
*<i>

s>

where

0=

XQ -,
y<>

being referred

to as the point

$ on the conic V.
x., p. 211.

* See Quarterly Journal of Mathematics, vol.

412

Transformations
a =
]3

the quadratic becomes (XQ - ay ) i.e. the square of a factor of the first degree also (1) reduces to x - ay + a?z = 0,

When

which is plainly the equation of the tangent at the point a to the conic V\ whence the line corresponding to a quadratic witli distinct roots is a chord of the conic F, this line becoming a
tangent when the roots are equal.

The only invariant that a quadratic has is its discriminant, and this is also an invariant in the ternary system, its vanishing
being the condition that the line corresponding to the quadratic now consider the system of two should touch the conic V.

We

quadratics

which

for shortness

we

call

and M.

When

transformed these become two lines

Jj

~ ax + by +

cz,

M = dx

+ b'y + cz.

Now the condition that the line whose equation is A L +

should touch the conic

V is
a'c

X 2 (ac -

2
ft

+ X M (ac +

2bb')

+ M 2 (aV -

2
ft' )

0.

(2)

All the coefficients of this equation are invariants in both systems we have already seen that this is true of the first and
:

last coefficients,

and the intermediate

coefficient
is

which

is

the

harmonic invariant of the binary system

an invariant in the

ternary system also, its vanishing expressing the condition that the lines Z), should be conjugate with regard to the conic F.

This equation determines the tangents which can be drawn

to the conic V. through the point of intersection of L and When this point is on the conic the tangents coincide, and the

discriminant of the quadratic vanishes. Whence we obtain geometrically the following form for the resultant of two quadratics
:

R - 4(ac for if Z,

V)

(a'c

2
ft'

(ac'

+ ac - 2bbJ

My and Fhave a common

point, the original

quadra-

Transformation of Binary
ties imist

to

Ternary Forms.
is

413

have a common

root,

and the condition

in each case

the same.

Again, the pairs of points or lines given by the equation XL + nM= form a system in involution (cf Ex. 2, p. 398), the double points or lines being determined by the equation (2)
.

and in the ternary system the corresponding pencil of lines passing through a fixed point determines on a conic a system of
points in involution, the double points being the points of contact of tangents

drawn

to the conic

from the fixed point.

If

we

consider next the three quadratics

seen that the determinant (i 2 c 3 ) is an invariant in both systems, its vanishing being the condition in the binary system that the quadratics should form an involution (Ex. 2, p. 398),
it is

and in the ternary system that the three corresponding lines should meet in a point.
a final illustration, we consider a system of three quadratics connected in pairs by the harmonic relations
a\c<i

As

+ a 2 Ci - 2bib =

0,

&c.

Transforming the quadratics, we obtain three lines X, Y, Z, which form a self-conjugate triangle with regard to the conic V.
relating to three mutually harmonic quadratics, that their viz., squares are connected by an identical linear relation (see Ex. 5, p. 360, and Art. 166), is suggested by a

The theorem

well-known property of conies X, Y, Z is of the form

for

expressed in terms of

whence, restoring the original variables

tically,

rr

vanishes iden-

and X, Y, Z become the original quadratics, each divided which may be seen to be the square root of its disfactor a by criminant (see (1), Ex. 5, p. 360).

414
186.

Transforma tions

The

tttuartic

and

metrically.

It will appear

Covariants treated geofrom the remarks to be made in

its

the next Article that in

under consideration

to the quartic

applying the transformation = (a, b, c d, e)(xQ ^o) 4

now
>

the

term 6cx

will be replaced

will be replaced

by 2cxz + cy by the two following conies

,

so that the quartic

:

2 If ~ ax

V^y*
the form of

+ cy z + ez z + 2dyz + 2czx + %bxy = -4zx = 0;

0,

here selected being connected with by an The invariants of and invariant relation. are invariants

of the original binary form, for the discriminant of

U - p V is

4 o3
/

-Ip +

J,

and the invariants

A' =

of the ternary system are

-4,

0' =

0,

7,

A =

<7;

where

/ and J are

the invariants of the quartic, and the dis-

criminant of

U - p V is written as usual under the form

A - pQ +
p'9'

/o

A'.

Let the conies

U and V intersect
by

these points being determined

in the points A, B, C, the equations

when
;

has the four values

a, /3, 7,

<5,

quartic and let the points J9<7, AD-, CA, #Z); AS,
triangle

of intersection of the

the roots of the binary common chords

be E, F, G, respectively, the being self-conjugate with regard to both conies. Now, denoting by (a)3) = the equation of the line AS, and using a similar notation for the remaining chords, we have by

CD

EFG

the theory of conies

where p^ p z

/o 3

are the roots of the equation

4/o

Ip +

J=

0.

The Quartic and its Covariants treated geometrically.

,

415

On restoring the original variables # y in these equations, vanishes Fo identically, and we have J7 resolved into a pair of quadratic factors in three different ways, depending on the
solution
of

the reducing cubic of the quartic.

"Whence

it

appears that the resolution of a quartic into its pairs of quadratic factors, and the determination of the pairs of lines which
pass through the intersections of two conies, are identical problems, each depending on the solution of the same cubic equation.

We now proceed to show that the sides of the common selfconjugate triangle of V, V correspond to the quadratic factors
of the sextic covariant in the binary system.

Since the side

are found by is the polar of the co-ordinates x, y of = 0, (aS) = we have, therefore, solving the equations (fly)
",

FG

y'

and, substituting for

polar of

#', y', z

the values thus determined in the

E,

viz.,

we

express this equation in the form

On restoring the original variables #, y^ this is seen to be one of the quadratic factors of the sextic covariant (see Art. It is therefore proved that the points where FG meets 166).

V are determined by
and consequently the

the quadratic equation

six points

on

V which

correspond to the

roots of the sextic covariant

are the points where this conic meets the sides of the common self-conjugate triangle of

and F.

To determine
of the Hessian,

the points on

F which correspond to the roots

7 and

we

calculate for the conies

Fthe

co-

416
variant
finding
conic

Tr ansformations.
JP (Salmon's Conic
Sections,

Art.

378)

thus

f (ae

- 2bd +

2
)

sir

+ (ad -

be) xy

and on restoring the

original variables,

we have

also, since the conic

intersects If

and

tact of their

common

tangents,

we

see that the points

F in the points of conon V

corresponding to the roots of the Hessian are the points so determined. The Hessian has, moreover, a double geometric origin, for it may equally well be obtained by transforming the
conic

(Salmon's Conies, Art. 377) which is the envelope of a and V. line cut harmonically by the conies

187.

When

the transformation of Art. 185,

viz.,

x =
applied to a quantic / (<rn 2/0) of even degree 2m, it is plain that the roots of this quantic will be represented geometrically
is
,

by the points

of intersection of a curve of the

th

degree with

If the degree of the quantic is odd, it must be squared before the transformation is effected and the roots will then be represented geometrically by the points of
;

the conic section F.

contact of the corresponding curve with the conic. In transforming the quantic /(# 2/o)> we may obtain an
>

indefinite

number
;

of ternary
if

transformation

for

forms by varying the mode of denote any one of these forms,

U+

0m-2 F,

in which the coefficients of m _ 2 are arbitrary, would equally well be a transformation of /(a? #<>), since this form would on restoring the original variables return to the quantic
>

/ (#o>

Moreover, every possible transformation is included in the foregoing. Among these innumerable ternary forms
yo)

there

is one, and only one, such that its invariants and covariants are invariants and co variants of the binary quantic

The Quartic and its Covariants treated geometrically

also. let
II

417

To determine this

form, take the tangential form of F, and

be the operator obtained by substituting x Dv Dz for the variables therein ; operating then with n on + ^ m _ 2 F, we - 2 and obtain a result w,_2 of the degree m equating to zero its coefficients, we have equations sufficient to determine all the
, ,

D U

coefficients of

m _3

The

required transformation
first

is

therefore

unique, as these equations are of the

degree.
to

The following method may be employed

proper form of

obtain the

corresponding to a given binary quantic

of even degree. Let the quartic be written in the form

u =

(#

i,

#2,

3,

#*) (#o>

^o)

3A 4

a^r

o 2

'

^u

\j

.7

dxtdy,

dy*}

transforming the second differential coefficients, and multiplying the terms by a?, y, 2, respectively, we obtain the proper form for U, such that D (U) = 0, viz.,

Again, in the case of the sextic

u,

writing

it

in the

form

d zu
-r

transforming the quartics

just explained,

=, 2

d*u
r
,

d zu

ofco

dx dy

35
dyf
s,

in

the

manner

and multiplying by

a?,

y,

respectively,

we

obtain a ternary cubic

manner the transformation

on that of the
;

In a similar of the proper form. of the octavic is made to depend

and proceeding in this way step by step sextic transform any binary quantic u of even degree to a of half the degree, such that II ( U) =0. ternary quantic

we may

The following examples are given to illustrate the transformation explained in the present section.

418

Transfer ma tions
EXAMPLES.

1.

If /(*o, yo) becomes

U (x,
%<?,

y, 2)

by the transformation
2#
yo,

x=
prove in general that

y=

= yo 2

where n

is

the degree of f(x, y), and

if II
(

n( ET)
0,

*Z

U - -

"Whence, in particular,

U)

prove that

where

*o, yo

and

*</,

yo'

are cogredient variables.

2. If

prove that
Since

n(^i +

y<^z

s^s)

when n (</>i) =

0,

n (4> 2 =
)

0,

n (4> 3 =
)

0.

dy<?

dzodyo

dxj'
d^s dx

we

have,

by Ex.

1,

fyi
dz

d<\n

'

dy

but

and therefore vanishes by what precedes.

statement at the end of Art. 187.

We

have thus a formal proof of the

"When

n(<i), n(4> 2 ),

n(<J> 3 )

do not vanish,

we have

in general

3. If two quantics M and w be transformed prove that the Jacobian of w, w in the binary system becomes the Jacobian of U, F, in the ternary system. Z Express /(M, w) in terms of XQ , zoyo, y<j* and the second differentials of u and
;

w, and then transform by Ex. 4. Prove that the quartics

(01 *
2

1.

+ 2/8i xy + 7iy2
+

(03

a;

+ 2)83^

+ 73y 3 )

(a2

+ Ifaxy + 72 y2
-f

2
)
,

(1)

(ai*

2o2 y + a3 y3 )

(7i

+ 272*y + 73 y2) -

(jS^

2faxy

+ )8 3 y

2 2
)

(2)

have the same invariants.

Examples.
Transforming
(01*
(2) to

419
the conic

the ternary system,

we have
732)

o2 y

032) (71*

+ yzy +
2
,

(fax

+ fay +

ft,*)

which

for shortness

we

write as

LN'- M

where
fat,

L=

aix

o2 y

+ o3 z,

If = fax

+ fay +

N = y\x + y y +
z

(3)

Now, when

the discriminant of

is

formed, the invariants of

(2) are

the functions
is

3H and G
removed).

of this cuhic in

(or

the last

two

coefficients

when

the second term

This discriminant

may

be obtained as the resultant of the three equations

NO.I

- 2Mfa + Lyi - 4x* = 0, Na* 2Mfa + Zyz + 2Xy = 0, JVos - 2 Mfa + Zy3 - 4\x = 0,
;

(4)

when

x, y, z are eliminated

or

by eliminating

by means of the three additional equations form

01

(3)

the six quantities x, y, z, L, M, the resultant is obtained in the

420
5.

Transformations.
Determine the condition that three quadratics should by linear transfor

mation be reducible to the forms

Am.
6.

/n/33 - 4/12/23 +

/22

2/22/31

0.

Prove that the condition in Ex. 5

:

is

the same for the following two sets of

quadratics

and

7.

Determine the condition that \u + vw should have two square

factors,

where u and

are quartics.

Transforming,

we have

in this case
z
;

\V+ n~P + vW = (ax + fry + yz)

consequently, every term in the tangential form of 2 2 2 fja>, v\, giving six equations to eliminate A. ju , j/
, ,

\U+ nV+vTP must vanish,

\/j.
;

hence the required con-

dition is determined.
8. If a quartic have a square factor, prove geometrically that this factor a quintuple factor of the covariant G> and construct the point on the conic which corresponds to the remaining root of the equation Gx = 0.
;

is

V
V

9.

where
10.
ratic

Resolve the quartic as in Art. 169 by finding the tangents to the conic meets it, J7and F having been expressed as sums of squares.

Express in terms of their invariants the resultant of the quartic and biquad-

11

Determine the condition that two quadratic factors

Z7b

(x

a)(x

),

(x

7) (xS)

of a quartic
volution.

should form with a given quadratic A* 2

2 fix

v a system in in-

is

Transforming, the three corresponding lines must meet in a point, which point one of the vertices of the common self-conjugate triangle of the conies Z7and V.
tangential equation of these points
is

The

J(2,

2',

*)

0,

which
2

is

therefore the

required condition, the tangential

12.

form of

nU + V

being

2 +

*c* -f 2'.

Apply the method of transformation of Art. 185

to prove the

theorem of

Art. 176.

Let Tschirnhausen's transformation be put under the form

Examples.
Make
replace z

421
;

the numerator and denominator of the last fraction homogeneous in x, y

by

- \, and

transform

(1)

becomes then

L + \L'=0,
where

L=

ax +

&>/

y:,

L'

= ax +

&y

-f

y'z.

If x, y, z be eliminated
shall

from the equations

;

L+

\L'

0,

U=

0,

V=

0,

we

have the transformed quartic in A. which, considered geometrically, determines the lines drawn from the point of intersection P of L and L' to the points of
intersection

A, B,

C,

D of U and

V.

Again,

if

K be so determined that the conic

U + /eFpass
PD,
is

through the point P, the anharmonic ratio of the lines PA, PB, PC, equal to the anharmonic ratio of the lines TA, AB, AC, AD, where TA is

the tangent to

U -f K V at A
t

that

is,

of the lines

Kt',

pit',

pzf,

pztf,

where t and t' are the tangents to Z7und V at A. Now, forming the invariants of the quartic whose roots are a, pi, pz , pi, the theorem follows by Arts. 170 and 177.
13.

Let three points

a, b, c

be taken on the conic px =

ai
2
<

V given

by the equations

-i-

b\

<J>

c\,

y, the roots of a cubic U; prove the fol<p at these points being a, lowing constructions for determining the points on the conic corresponding to the roots of the cubic covariant Gx and the Hessian x

the values of

1. Let tangents be drawn to the conic V at the points angle ABC, the lines Aa, Bb, Cc meet the conic at points a',
the roots of

a, b, c,
b', c

forming a

tri-

corresponding to

Gx

2. The four

triangles abc, a'b'c',

ABC, A'B'C'

are homologous, and their axis

of homology meets the conic

14.

V at the

points corresponding to the roots of

the constructions in the last example prove that U, and Ox have the same Hessian M , and that the roots of x are imaginary when the roots of U* are

From

real.

Dublin Exam. Papers, Bishop Law's Prize, 1879.

CHAPTER

XVII.

THE COMPLEX VARIABLE.

188.
tities.

Graphic Representation of Imaginary QuanThe imaginary expression a + b </- 1 may be written

(cos a + sin a

in the form

fj.

*/-

1),

where

z
,

and tan a =

-.

It

may

be regarded, therefore, as determined by the linear

;

magnitude

and the angle a ju being called the modulus, and ju, a the argument of the imaginary quantity. Let rectangular axes OX, (fig. 7) be taken; and a

OY

point

such

that

We
ju

have then cos a = a, and

fjL

OM =

AM
The

sin a

b.

expression a + b </- 1 may therefore be re-

presented graphically by the right line

drawn from

to a

Fig.

7.

point in a plane whose co-ordinates referred to the fixed axes of this point from the origin being are a, b ; the distance the modulus, and the angle equal to the arguequal to

OA

XOA

ment

of the

imaginary quantity.

Addition and Subtraction of Imaginaries.

423

The magnitude
the magnitude of vanishes (that is,

of

its

an imaginary quantity is estimated l>y modulus. When the imaginary quantity

vanishes

then

when a and b separately vanish) its modulus and, conversely, when the modulus vanishes, since + b* = 0, a and b must separately vanish, and therefore
;

the imaginary quantity itself. Two imaginary quantities, a + ib and a' + ib', are equal when a = a' and b = b', i. e. when the

moduli are equal and when the arguments

differ

either are equal or

by a multiple of 2ir. In what follows we shall

of a

for brevity represent the

modulus

and argument

^/-

by the notation
arg.
(a

mod. (a +

ib),

ib),

where

as usual represents

189.

^/- 1. Addition and Subtraction of Imaginaries.

a'

Let a second imaginary quantity

right line OA', so that

ib'

be represented by the

OA' = mod.

(a

ib'),

XOA' =
mode
4-

arg. (a +

ib').

We

proceed to determine the

a
4-

of representing the

sum

ib

a'

ib'.

Writing
in

this

sum

in the form a + a + i(b +

b'),

we

observe,

accordance with the convention of Art. 188, that it will be represented by the line drawn from the origin to the point

whose co-ordinates are a + a, b + b'. To find this point, draw AB parallel and equal to OA'; since AP, BP, are equal to a', b', B is the required point, and we have

OB=mod.

[a

a'

+ i(b+b')},

XOB = arg.{a + a

i(b

b')}.

To add two imaginary

;

quantities, therefore,

we draw

OA

to

to represent and, at its extremity, represent one of them the second (that is, so that its length is equal to the modulus,

AB

and

the angle
;

it

makes with

OX

second)

then

OB

represents the

equal to the argument, of the sum of the two imaginary

424
quantities.
the

The Complex Variable.

Since

OB

is less

than

OA

AB,

it

follows that

modulus of the sum of two imaginary quantities

is less

than the

sum of their moduli.

This mode of representation
of

may be extended

to the addition

any number
ib",

of

imaginary quantities.

Thus, to add a third

by OA", we draw BC parallel and equal to Then 00 represents the sum of the three imaginary quantities OA, OA', OA". It is evident also that we may conclude in general that the modulus of the sum of any number of imaginary quantities is less than the sum of their
a" +

represented

OA", and join 00.

moduli.

Subtraction of imaginaries can be represented in a similar way. Since OB represents the sum of OA and OA', OA will
represent the difference of

OB

and OA'.

To

subtract

two

imaginary quantities, therefore, we draw at the extremity of the line representing the first a line parallel and equal to the
second, but in an opposite direction
(i. e.

a direction which makes

first).

with

OX
join

an angle greater by

TT

than the argument of the

We

to the extremity of this line to find the right line

which represents the difference of the two given imaginaries. 190. Multiplication anil Division of Imaginaries. To multiply the two imaginary quantities a + ib, a' + ib', we
write

them
a+
ib

in the

form
+
i

=
JJL

(cos a

sin a),

a'

ib'

p' (cos a'

sin a').

We have then,
(a

by De Moivre's theorem,
+
ib')

ib) (a

= pp {cos (a +

a')

sin (a

a')

) ,

the product of two imaginary quantities is an imaginary quantity of the same form, whose modulus is the product of the two moduli, and whose argument is the sum of the two argu-

which proves that

ments.

In the same way

of
is

it

appears that the product of any number

imaginary factors is an imaginary quantity, whose modulus the product of all the moduli, and whose argument is the sum of all the arguments.

The Complex Variable.

425

To

divide a +

ib

by a +

ib',

we have
a')

similarly

(cos (a

sin (a

a')}
,

which proves that

the quotient of two imaginary quantities

is

an

imaginary quantity of the

same form,

tvhose
is

modulus

is

the quotient

of the two moduli, and whose argument

the difference of the two

arguments. It is evident from the foregoing propositions that any power of an imaginary quantity, e.g. (a + ib) m can be expressed in the
,

form

+ iB, where
if

A and B are

real quantities.

And, more

generally,

in

any polynomial

whose coefficients are either real or imaginary quantities, an imaginary quantity a + ib be substituted for the variable z, the result can be expressed in the standard form of imaginary quantities, viz.

+ iB.

was assumed in the proof of the theorem of Art. 16 that when a product of any number of factors (real or imaginary) This is evident when vanishes, one of the factors must vanish.
It

the factors are

all real.

From what

is

above proved the sa me

;

conclusion holds

the factors are imaginary for, in ord er that the modulus of the product may vanish, one of its factors

when

must vanish, and therefore the imaginary quantity of that factor is the modulus.

which

In the earlier Chapters of 191. The Complex Variable. the present work the variation of a polynomial was studied corresponding to the passage of the variable through real values
and the mode of representing by a figure the form of the polynomial was explained. Such a mode of treatment is only a particular case of a more general inquiry.
from
oo to

oo

Given a polynomial
f(z)

= aQ zn +

fliS"'

+ &2 s n 2 +

+ a n -\z + a n

we may study its variations

corresponding to the different values

426
of
2,

The Complex Variable.

where
z

both take

all possible real values.

has the imaginary form x + iy, and where x and y This form x + iy is called the

complex variable. All possible real values of the variable are of course included in the values of x + iy, being those values which
arise

by varying x and putting y =

0.

In accordance with the

the imaginary quantity

principles of Art. 188

we may represent

# +

to hy the line OP (fig. 8j drawn from a fixed origin the point whose co-ordinates are x, y. Or we may say, x + iy is Thus all possible values of x + iy represented by the point P.
iy

by all the points in a plane. Since for any + iB (Art. 190), the value of z,f(z) takes the form particular values of f(z) may be represented in a similar manner by points

will be represented

in a plane.

We confine
We

ourselves in the present Article to the

representation of the

variable

x +

iy itself.

conceive the

variation of x + iy to take place in a continuous manner ; for ex-

ample, by the motion of the

point
x, y,
f

along a curve.

If

OP
we
:

and

OP

represent two consecu-

tive values of the variable,

x+

write the corresponding values as follows x' + iy, iy,

Fig. 8.

z=x + iy = r (cos
Since
follows that
if z'

+ i sin

0),

z'

=x +
of

iy

= r (cos

0' + i sin 0').

f

OP' represents

the

sum

OP and PP

(Art. 189), it
;

PP'
h

h,

may

represents the imaginary increment of 2 be written in the form

and

h = p

(cos

sin 0),

where p =

PP,

and

is

the angle
of z is

PP' makes with OX.

z is

The

variation of the

modulus of

OP" or 0'

OP or r;

the

variation of the
of z itself
is

argument

P'OP

- 6

the variation

h or p (cos + i sin 0), as just explained. Let the point be supposed to describe a closed curve.

When
;

it

returns to

P the modulus takes again its original value

and

Continuity of a Function of x

iy.

427

the argument takes its original value if the point is exterior to the curve, or is increased by 2?r if is interior to the
curve.

site directions,

If the complex variable describes the same line in two oppothe variations of its argument are equal and of From this we opposite signs, i. e. the total variation is nothing.

can derive a property of the variation of the argument of the complex variable, which will be found of importance in our suc-

ceeding investigations. Let a plane area be divi-

ded into any number of parts

by
(fig.

lines
9)
;

UD, AF, EC,

&c.

then the variation

to

of the argument relatively

the perimeter

of the

ichole

area

is

equal

to the

sum of its
Fig.
9.

variations relatively to the pe-

rimeters of the partial areas :

all

the areas being supposed to be described by the variable moving in the same sense. This is evident for when the
;

point is made to describe all the partial areas in the same sense, each of the internal dividing lines will be described twice, the two descriptions being in opposite directions and the exter;

nal perimeter will be described once hence the total variation of the argument relatively to the dividing lines vanishes, and
;

the variation relatively to the external perimeter alone remains. In the figure, for example, when the point describes the areas in the sense indicated by the arrows, the total ABF,

AFD

variation relatively to the line

AF vanishes.
Complex

192.

Continuity of a

Function of the

Variable.
a fixed h = (cos

Suppose the complex variable z, starting from value z to receive a small imaginary increment
,

sin

r/,)

we have

then,

if

f(s)

be the given

function,
h)

&c.,

428
and the increment

The Complex Variable.

of /(*), being equal to/(*

h) -/(so), is

1.2

1.2.3

In this expression the coefficients of the powers of h are all imaginary expressions of the usual form and if their moduli be z 3 #, b, c, &c., the moduli of the successive terms are ap, bp cp &c. ;
;

and

since,

by

Art. 189, the modulus of a

it

sum

is

less

than the

sum

of the moduli,
is less

follows that the

modulus of the increment

of /(s)

than
z 3 ap + bp + cp + &c.

Now

a value

may

or any less value of

be assigned to p (Art. 4), such that for it, the value of this expression will be less /o,

than any assigned quantity. It follows that to an infinitely small variation of the complex variable corresponds an infinitely small variation of the function in other words, the function varies
;

continuously at the same time as the complex variable

itself.

Variation of the Argument of f(z) corresponding to the Description of a small Closed Curve by the Complex. Variable. Corresponding to a continuous series of values of z we have a continuous series of values of f(z) which can be
193.
9

represented, like the values of s itself, by points in a plane. represent these series of points by two figures (fig. 10) side

We

Fig. 10.

by

side,

on different planes.

which, to avoid confusion, may be supposed to be drawn To each point P, representing x + iy, cor-

Variation of the

Argument off(z\

c.

429

responds one determinate point

describes a continuous curve,

P representing P describes also

/(s).

When P

a continuous

returns to its original position after describcurve; and when returns also to its original position. ing a closed curve,

Our present object

by P.
#o,
l/o 9
:

is

to discuss the variation of the

argument

of /i's) corresponding to the description of a small closed curve

Let
i-

A
=

e. s

be any determinate point whose co-ordinates are divide the discussion into two #0 + iyv

We

cases

(1).

When
is

y +

iyQ is

not a root of /(s) =

zero.
0,

0, i.e.

when

/(s

()

different

from

(2).

When
In the

x*+

iy c is

a root of/(s) to the point

),

or/(s

=
)

0.

(1).

first case,

A
;

corresponds a point

A
P

representing the value of f(z Let s = s + />, where h =

represents
s,

to

and OfA! is different from zero. + i sin (cos 0) and suppose P, which p describe a small closed curve round A. Let
then

represents the increment of f(z) correpresent /(s) ; of z. By the previous Article responding to the increment

A'P

AP

it

modulus
less

appears that values so small may be assigned to /o, that the of the increment of /(s), namely A'P, may be always

than the assigned quantity 0'A\ hence

P may be supposed

a closed curve so small that the correspondwill be exterior to 0'. It foling closed curve described by lows, by Art. 191, that corresponding to the description by of a
to describe

round

small closed curve, which does not contain a point satisfying the
equation f(z) = 0, the total variation of the argument of f(z)
nothing.
(2).
is

In the second
repeated

case,

suppose X Q + iyQ

is

a root of the equa-

tion f(z) =

times,

and

let

/()-(-,)**();
then
f(z)

= hm

\f,(z)

= pm
;

(cos

w0 +

sin

m^)

;//

(z).

In this round A,

case O'A'

=
its

and when

P describes a closed

curve

P returns to

original position,

and the argument

430

The Complex Variable.

by a multiple
of
27r,

of f(z) will be increased

which

may

be de-

termined as follows

From

the above equation

we have

and the increment of arg. f(z) will be obtained by adding the increment of mfy to the increment of arg. $ (z) Now the latter increment is nothing by (1), since the curve described by P may and since the increbe supposed to contain no root of if (s) =
.

ment
2tmr.

of

is 2?r

in one revolution of P, the increment of

m$

is

It follows that when

describes a small closed curve con-

taining a root of the equation f(z)

0, repeated

times, the argu-

ment off(z)
194.

is

increased by 2mir.

Cauchy's Theorem.
two opposite
its

When

z describes

the same

line in a plane in

directions, f(z) describes the cor-

responding line in
arg. f(z)

plane in two opposite directions, and the

that

if

undergoes equal and opposite variations. It follows any plane area be divided into parts, as in Art. 191, the

variation of the arg. f(z) corresponding to the description in the same sense by z of all the partial areas, is equal to the varia-

tion of arg. f(z) corresponding to the description by z of the let any closed perimeter in the external perimeter only.

Now

plane

XY be described

and suppose, in the

first

place, that it

It can contains no point which satisfies the equation /(s) = 0. be broken up into a number of small areas, with respect to each

of which the conclusions of (1) Art. 193 hold; been just proved it follows that the
to the description

and by what has

variation of arg. f(z) corresponding by z of the closed

perimeter is nothing. Suppose, in the second place, that the closed perime-

which is a root = of the equation /(z) repeated m Let a small closed curve times.
ter contains a point

PQRS
point.

be

described

round

this

Fig 1L
-

The

tion

by

& of

variation of arg.f(z), corresponding to the descripthe whole perimeter, is equal to the sum of its

Number of Roots of the General

variations

Equation.
of

corresponding

to

the

description

the

areas

ABCPSR, CDARQP, PQRS.

;

The two former

variations

vanish by what is above proved and the latter is, by (2), The total variation, therefore, of /(z) Art. 193, equal to 2tmr. is 2/W7T. Similarly, if the area includes a second, third, &c.,
points which represent roots repeated m f + m" + &c.) TT. total variation = 2 (m +
f
,

Hence we

m"> &c., times, the derive the

following theorem due to Cauchy

The number of roots of any polynomial, comprised within a given

plane area,
is

obtained by dividing by 2?r the total variation of the

argument of
195.

this

polynomial corresponding

to the

complete descrip

tion by the com/itcx variable of the perimeter of the area.

Number of Roots of the General Equation.

"We are enabled by means of the principles established in the preceding Articles to prove the theorem contained in Arts. 15 and 16 namely, every rational and integral equation of the n th
;

degree has n roots real or imaginary.

Let
/(z)

=a

+ aiZ n

~l

+ a 2 zn

~z

a^z + an

be a rational and integral function of

z.

Without making any

= further than supposition as to the existence of roots of f(z) that/(z) cannot vanish for any infinite values of the variable,

we can suppose
no root

z to describe in its
it.

plane a circle so large that

exists outside of
/(z)

If, then,

z n {a

#iz'

+ a2 z'* + where

+ a n z'n
-, z

= whose modulus
is

z"0(s'),

z'

z' 9

the reciprocal of the modulus of

z,

will

describe a small circle containing a portion of the plane corresponding to the part outside of the circle described by z and
;

no root of ^ (z') =

will be included within this small circle.

circle

Hence, corresponding to the description of the whole the variation of arg. ^ (*') = 0> an(*, therefore,
variation of arg. /(z) = variation of arg. z n

by z,

432
and
is if

The Complex Variable.

z

= r (cos 9 +

n sin 0), or z = r n (cos

nO + i
n
is

sin nO) ,

increased

by

27r,

and, therefore, arg. z

increased

by

2mr.
It follows

from Cauchy's theorem, Art. 194, that the number

of roots comprised within the circle described number of roots of the equation f(z) = 0, is n
is

by z, i. e. the total and the theorem

;

proved.

The proposition whose proof was deferred in Art. 15 is thus shown to be an immediate consequence of Cauchy's theorem, which may therefore be regarded as the fundamental proposiTheory of Equations. It is proper to observe, howtheorem of Art. 15, viz., that every equation has a root, can be proved directly, and independently of Cauchy's theorem, by aid of the principles contained in Art. 193 and the
tion of the
ever, that the

preceding Articles, as

we proceed now
>

to show.

If possible, let there be no value of z which makes /(s) vanish; and let the value s represented by A, fig. 10, correto the origin 0'. spond to the nearest possible position, A', of

Now,

giving

a small increment

/?,

and considering the

first

term/'(s )A of the corresponding increment of /(s ), it is seen that the directions in which these two small increments take
place are inclined at a constant angle. It is possible therefore, by properly selecting the direction of the increment h, to cause
the increment of f(z ) to take place in the direction A'Cf, and thus to make A' approach nearer to the origin, which is contrary to hypothesis.
It follows that the

minimum
zero,

value of the

modulus

of f(z) cannot be different

from

and therefore that

some value of z exists which makes / (z) vanish. In note D will be found some further observations on the
subject of this Article.

NOTE
NOTE
THE

S,

A.

ALGEBRAIC SOLUTION OF EQUATIONS.

solution of the quadratic equation

was known
treatise

to the Arabians,

aii-1 is

found in the works of ^Mohammed Ben Musa and other writers

published in the ninth century.

In a

on Algebra by Omar
with methods
of

Alkhayyami, which belongs probably

century,
is

to the middle of the eleventh

found a

classification of cubic equations,

;

but no attempt at a general solution. The of was introduced into Italy from the Arabian writers Algebra study of in Pisa the thirteenth century and for a long Leonardo early by
geometrical construction
;

period the Italians were the chief cultivators of the science. work, styled ISArte Maggiore, by Lucas Paciolus (^known as Lucas de Burgo) was published in 1494. This writer adopts the Arabic classification
of cubic equations, and pronounces their solution to be as impossible in the existing state of the science as the quadrature of the circle. At the same time he signalizes this solution as the problem to which the

attention of mathematicians should be next directed in the development of the science. The solution of the equation y? + mx = n was
effected

by Scipio Ferreo
it

but nothing more

is

known

of his discovery

than that he imparted

attention of Tartaglia

to his pupil Florido in the year 1505.

The

was directed to the problem

in the year 1530, in

consequence of a question proposed to him by Colla, whose solution depended on that of a cubic of the form 3? + par = q. Florido, learning
that Tartaglia had obtained a solution of this equation, proclaimed his own knowledge of the solution of the form a? + mx = n. Tartaglia,

doubting the truth of his statement, challenged him to a disputation

2*

434
in the year 1535
;

Notes.
and in the mean time himself discovered the solu+ mx = w. This solution depends on assuming
of the difference of

tion of Ferreo's form y?

for

x an expression l/t - ^/u consisting

;

two

radi-

and, in fact, constitutes the solution usually known as Cardan's. Tartaglia continued his labours, and discovered rules for the solution
cals of the various

forms of cubics included under the classification of the

Cardan, anxious to obtain a knowledge of these rules,

Arabic writers.

applied to Tartaglia in the year 1539; but without success.

After

many
rules
;

solicitations Tartaglia

imparted to

him

a knowledge of these

receiving from him, however, the most solemn and sacred pro-

mises of secrecy.

Regardless of his promises, Cardan published in

1545 Tartaglia' s rules in his great work styled Ars Magna. It had been the intention of Tartaglia to publish his rules in a work of his

own.

He commenced

the publication of this

work

in 1556

but died

in 1559, before he

had reached the consideration of cubic equations. As his work, therefore, contained no mention of his own rules, these rules came in process of time to be regarded as the discovery of Cardan, and to be called by his name.

The

solution

of equations of

the fourth degree was the next

;

problem to engage the attention of algebraists

and here,

as well as in

the case of the cubic, the impulse was given by Colla, who proposed 4 2 to the learned the solution of the equation # + 6# + 36 = 60#. Cardan

this

appears to have made attempts to obtain a formula for equations of kind but the discovery was reserved for his pupil Ferrari. The
;

method employed by Ferrari was the introduction

in such a

of a

new

variable,
;

way
is,

as to

make both

sides of the equation perfect squares

this variable itself being

gree.
is

It

in fact,

determined by an equation of the third deThis solution virtually the method of Art. 63.
solution

sometimes ascribed to Bombelli,

Algebra, in 1579.
lished

The

who published it in his treatise on known as Simpson's, which was pubno respect essentially different
treatise,

much

later (about 1740), is in

from that
in

of Ferrari.

In the year 1637 appeared Descartes'

which

are found

many improvements
of Signs."

in algebraical science, the

chief of

which

are his recognition of the negative

and imaginary roots

of the biqua-

of equations,

and his " Eule

His expression

two quadratic factors, although deducible Ferrari's from form, was an important contribution to immediately Euler's algebra was published in 1770. the study of this quantic.
dratic as the product of

His solution of the biquadratic (see Art. 61)

is

important, inasmuch

Nolc*.
as it brings the treatment of this

435

cubic

form into harmony with that of the by moans of the assumed irrational form of the root. The methods of Descartes and Euler were the result of attempts made to
obtain a general algebraic solution of equations.

Throughout the

eighteenth century
this

problem

occupied themselves with but their labours were unsuccessful in the case of

many mathematicians

equations of a degree higher than the fourth. In the solutions of the cubic and biquadratic obtained by the older the first, illusanalysts we observe two distinct methods in operation
:

trated

by the assumptions of Tartaglia and Euler, proceeding from an assumed explicit irrational form of the root the other, seeking by the
;

aid of a transformation of the given function, to change its factorial

character, so as to reduce it to a form readily resolvable.

In Art. 55

these two methods are illustrated

tion of

which

is

together with a third, the concepto be traced to Yandermonde and Lagrange, who pub;

lished their researches about the

1771.

The former

of these writers

same time, in the years 1770 and was the first to indicate clearly

the necessary character of an algebraical solution of any equation, viz., that it must, by the combination of radical signs involved in it,
represent any root indifferently when the symmetric functions of the roots are substituted for the functions of the coefficients involved in

the formula (see Art. 94). His attempts to construct formulas of this character were successful in the cases of the cubic and biquadratic ;

but failed in the case of the quintic.

Lagrange undertook a review of

the labours of his predecessors in the direction of the general solution of equations, and traced all their results to one uniform principle. This
of the given equation to principle consists in reducing the solution that of an equation of lower degree, whose roots are linear functions

of the roots of the given equation

and

of the roots of unity.

He shows

also that the reduction of a quintic cannot

be effected in this way, the

having whether any such solution was possible at all. Demonstrations have been given by Abel and Wantzel (see Serret's Cours tfAlgebre Supeof resolving algebraically equars, Art. 516) of the impossibility tions unrestricted in form, of a degree higher than the fourth.

of the sixth degree. equation on which its solution depends being All attempts at the solution of equations of the fifth degree should inquire failed, it was natural that mathematicians

transcendental solution, however,

of the quintic has

elliptic

M. Hermitc,

in a

form involving

been given by integrals. Among other

436

Notes.

contributions to the discussion of the quintic since the researches of Lagrange, one of leading importance is its expression in a trinomial

form by means of the Tschirnhausen transformation (see Art. 179). Tschirnhausen himself had succeeded in the year 1683, by means of
the assumption y =
quartic, and
in the reduction of the cubic and had imagined that a similar process might be applied to
,

P+

Qx + xz

the general equation.

The reduction

of the quintic to the trinomial

form was published by Mr. Jerrard in his Mathematical Researches, 1832-1835 and has been pronounced by M. Hermite to be the most
;

important advance in the discussion of this quantic since Abel's demonstration of the impossibility of its solution by radicals. In a Paper published by the Rev. Robert Harley in the Quarterly Journal
of Mathematics, vol. previously effected,
vi. p. 38, it is

shown that

this reduction

had been

by a Swedish mathematician named with Of Bring's reduction is Dr. Sylvester's Bring. equal importance transformation (Art. 180), by means of which the quintic is expressed
in 1786, as the sum of three fifth powers, a form which gives great facility to Other contributions which have been the treatment of this quantic. made in recent years towards the discussion of quantics of the fifth

riants of these forms.

is

and higher degrees have reference chiefly to the invariants and covaFor an account of these researches the student
referred to Clebsch's Theorie der binaren algelraischen Formen, and

to Salmon's Lessons Introductory to the

Modern Higher Algebra. There has also grown up in recent years a very wide field

of in-

vestigation relative to the algebraic solution of equations, known as the " Theory of Substitutions." This theory arose out of the researches of

Lagrange before referred

to,

and has received large additions from the

labours of Cauchy, Abel, Galois, and other writers. Many important but the subject is results have been arrived at by these investigators of too great extent and difficulty to find any place in the present
;

work.

The reader
1

desirous of information on this subject

is

referred

to Serret's Cours d" Algebre Superieure,

tions et des Equations Algebriques,

and

to the Traite des Substitu-

by M. Camille Jordan.

Xutes.

437

NOTE

B.

SOLUTION OF NUMERICAL EQUATIONS.

The

first

rical equation*

attempt at a general solution by approximation of numewas published in the year 1600, by Yieta. Cardan
of

had previously applied the rule

regula aurea")
to the

"false position" (called by

him

method were

of little

cubic; but the results obtained by this value. It occurred to Yieta that a particular

numerical root of a given equation might be obtained by a process analogous to the ordinary processes of extraction of square and cube
roots; and he inquired in what way these known processes should be modified in order to afford a root of an equation whose coefficients are i-iven numbers. Taking the equation / (#) = Q, where Q is a given

number, and
numerical

/ (a:)

a polynomial containing different powers of

.r,

with

Vieta showed that, by substituting in / (#) a known approximate value of the root, another figure of the root When this expressed as a decimal) might be obtained by division.
coefficients,

figure of the root

of this

value was obtained, a repetition of the process furnished, the next and so on. It will be observed that the principle
;

method
of

is

identical with the

main principle involved in the

approximation of Newton and Horner (Arts. 100, 101). All that has been added since Vieta' s time to this mode of solution of

methods

numerical equations
afford facility

How
of

is the arrangement of the calculation so as to and security in the process of evolution of the root. great has been the improvement in this respect may be judged

p. 603,

in Montucla's Histoire des Mathematiques, vol. i. where, speaking of Vieta's mode of approximation, the author regards the calculation (performed by Wallis) of the root of a biquadratic to eleven decimal places as a work of the most extra-

by an observation

vagant labour.
great ease

The same calculation can now be conducted with by anyone who has mastered Homer's process explained in
of

the text.
approximation was published in 1669; but before this period the method of Vieta had been employed and simAfter the period of plified by Harriot, Oughtred, Pell, and others.

Newton's method

Xewton, Simpson and the Bernoullis occupied themselves with the

4:)8

Xotes.

same problem. Daniel Bernoulli expressed a root of an equation in the form of a recurring series, and a similar expression was given by Euler; but both these methods of solution have been shown by
Lagrange
to be in no respect essentially different from Newton's solution (Traite de la Resolution des Equations Numeriques}. Up to

the period of Lagrange, therefore, there was in existence only one distinct method of approximation to the root of a numerical equation and this method, as finally perfected by Homer, in 1819, remains at
;

the present time the best practical method yet discovered for this purpose.

Lagrange, in the work above referred

in the methods of Yieta and

to,

Newton.

With
trials
;

pointed out the defects reference to the former

it

he observed that

it

required too

many
As

and that

could not be

depended
signalized

on, except

when

all

the terms on the left-hand side of the

defects in

equation /(#) =
first,

Q were

positive.

Newton's method he

its failure to

give a commensurable root in finite

terms

secondly, the insecurity of the process

;

which leaves doubtful

the failure of the

the exactness of each fresh correction

and

lastly,

method in the

case

of

an equation with roots nearly equal.

The

problem Lagrange proposed to himself was the following: "Etant donnee une equation numerique sans aucune notion prealable de la
grandeur ni de 1'espece de ses racines, trouver la valeur numeriquo exacte, s'il est possible, ou aussi approchee qu'on voudra de chacune
de ses racines."
Before giving an account of his attempted solution of this problem, is necessary to review what had been already done in this direction,

it

in addition to the methods of approximation above described. Harriot discovered in 1631 the composition of an equation as a product of
factors,

and the relations between the roots and coefficients. Viet a had already observed this relation in the case of a cubic but he This failed to draw the conclusion in its generality, as Harriot did.
;

discovery was important, for it led to the observation that any integral root must be a factor of the absolute term of an equation, and Newton's

Method

of

Divisors for the determination of such roots was

a natural result.

Attention was next directed towards finding limits

of the roots, in order to diminish the labour necessary in applying the

method

of divisors as well as the

methods

of approximation previously

in existence.

Descartes, as already remarked,

was the
;

first to

recog-

nise the negative

and imaginary roots

of equations

and the inquiry

Xotes.

439
number
of real

commenced by him
Stirling,

as to the determination of the

and

of imaginary roots of

any given equation was continued by Newton,

at a solution of the

De Gua, and others. Lagrange observed that, in order to arrive

it

problem above stated,

another.

was

first

necessary to determine the

of the real roots of the given equation,

and

For

this purpose

he proposed to

number them one from employ the equation whose

to separate

roots are the squares of the differences of the roots of the given equation. "Waring had previously, in 1762, indicated this method of

Note

separating the roots; but Lagrange observes (Equations Numeriques, iii.), that he was not aware of Waring' s researches when he

composed his own memoir on

the equation of differences
is

this subject.

It is evident that
possible,

when

formed,

it is

by rinding an
less

inferior limit to its positive roots, to obtain a

number

than the
substi-

least difference of the real roots of the given equation.

By

tuting in succession numbers differing by this quantity, the real roots of the given equation will be separated. When the roots are separated in this way Lagrange proposed to determine each of them by

the method of continued fractions, explained in the text (Art. 105). This mode of obtaining the roots escapes the objections above stated
to

Newton's method, inasmuch

is

as the
;

amount
is

of error in each sucis

cessive approximation

known

and when the root

commensurable

the process ceases of

itself,

and the root

of

given in a finite form.

Lagrange gave methods also equations, and observed that

could be obtained in the
(see Art. 74).
first

obtaining the imaginary roots of the equation had equal roots they instance by methods already in existence
if

Theoretically, therefore, Lagrange's solution of the problem

which
it

he proposed to himself
is

is

perfect.

As a

practical method, however,

almost useless.

The formation

of the equation of differences for

equations of even the fourth degree is very laborious, and for equabecomes well nigh impracticable. Even if the more convenient modes of separating the roots discovered sim-itions of higher degrees

Lagrange's time be taken in conjunction with the rest of his

still this

pro<

open to the objection that it gives the root in the form of a continued fraction, and that the labour of obtaining
process
is
it

in this

form

is

by Homer's process

greater than the corresponding labour of obtaining it It will be observed in the form of a decimal.

also that the latter process, in the perfected

form

to

which llorntr

440
has brought above stated.
it,

Notes.
is free

from

all

the objections to Newton's method

Since the period of Lagrange, the most important contributions to the analysis of numerical equations, in addition to Homer's improvement of the method of approximation of Vieta and Newton, are those The researches of Budan were pubof Fourier, Eudan, and Sturm.
lished in 1807
is
;

and those

of Fourier in 1831, after his death.

There

no doubt, however, that Fourier had discovered before the publication of Sudan's work the theorem which is ascribed^to them conjointly
in the text.

The

researches of

methods

of separation of the roots proposed

explained in Chapter IX. that of Horner, we have

The by these writers are fully By a combination of these methods with

a solution of Lagrange' s problem far

Sturm were published in 1835.

now

And it appears simpler than that proposed by Lagrange himself. In impossible to reach much greater simplicity in this direction.
extracting a root of an equation, just as in extracting an ordinary square or cube root, labour cannot be avoided ; and Horner' s process

The separation of the more are nearly equal, must remain roots also, especially when two This labour may admit of some reducn work of more or less labour.
appears to reduce this labour to a minimum.
or tion

by the
so

consideration of the functions of the coefficients

which

important a part in the theory of the different quantics. If, play for example, the functions H, /, and /", are calculated for a given
quartic, it will be possible at once to tell the character of the roots Mathematicians may also invent in process of time (see Art. 68).

some mode

of calculation applicable to "numerical equations analogous

to the logarithmic calculation of simple roots.

But

at the present
is to

time the most perfect solution of Lagrange's problem in a combination of the methods of Sturm and Horner.

be sought

Notes.

441

NOTE

C.

DETERMINANTS.
The expressions which form the subject-matter of Chapter XI. first called "determinants" by Cauchy, this name being adopted by him from the wiitings of Gauss, who had applied it to certain
were
.-pecial clashes of these functions, viz. the discriminants of binary and Although Leibnitz had observed in 1693 ternary quadratic forms. the peculiarity of the expressions which arise from the solution of linear equations, no further advance in the subject took place until

Cramer, in 1750, was led to the study of such functions in connexion with the analysis of curves. To Cramer is due the rule of signs
of Art. 108.

subject

During the latter part of the eighteenth century the was further enlarged by the labours of Bezout, Laplace,

Vandermonde, and Lagrange. In the present century the earliest cultivators of this branch of mathematics were Gauss and Cauchy;
the former of

whom,

in addition to his investigations relative to the

discriminants of quadratic forms, proved, for the particular cases of the second and third order, that the product of two determinants
is

itself

a determinant.

To Cauchy we

are indebted for the first

In his memoir on Alternate Functions, formal treatise on the subject. de V Kcole Poll/technique, vol. x., he disin Journal the published
cusses

determinants
several

as

a particular class

of

such functions, and

proves
ATeut

impulse was given

A important general to the study of these expressions by the

theorems relating to them.

writings of Jacobi in Crelle's Journal, and by his memoirs published in 1841. Among more recent mathematicians who have advanced
this

subject

may

be

mentioned

Caylcy, Sylvester, mathematics, pure or applied, in which the employment of this calculus is not of great assistance, not only furnishing brevity and elegance in the demonstration of known properties, but even leading
to

and Salmon.

Hermite, Hesse, Joachimsthal, There is now no department of

new

discoveries in mathematical science.

Among
to

recent works

which have rendered

L'-ndon, 1851

this subject accessible to students

may be menJhtermimmts, 1'avia, 1854

tioned Spottiswoode's Elementary Theorems relating

;

Brioschi's

La

teorica dei Jh'fi'rminanti,

Baltzer's Theorie
1) Bator's

und Anwendung der Determinants, Leip/ig, 1864

la theoric des

Elements de

Determinants, Paris, 1877

Scott's

442

Notes.

Theory of Determinants, Cambridge, 1880; and the chapters in Salmon's Lessons introductory to the Modern Higher Algebra, Dublin, 1876. For further information on the history of this subject, as well as on
that of Eliminants, Invariants, Covariants, and Linear Transformations,

the reader

is

referred to the notes at the end of the

work

last

mentioned.

NOTE

D.

THE PROPOSITION THAT EVERY EQUATION HAS A ROOT.

It
is

and what
the

important to have a clear conception of what is established, it is possible to establish, in connexion with the proposition
If in the equation
. . .

discussed in Art. 195.

coefficients a
,

a&? +

#B

-1

an =

a lt

without any restriction denote numbers, either real, or complex numbers of the form treated in Chapter XVII., then, with reference to such an equation it is not
;

a n are used as mere algebraical symbols that is to say, if they are not restricted to

proved, and there exists no proof, that every equation has a root. The proposition which is capable of proof is that, in the case of any rational integral equation of the n th degree, whose coefficients are all

complex (including real) numbers, there exist n complex numbers which satisfy this equation; so that, using the terms number and numerical in the wide sense of Chapter XVII., the proposition under consideration might be more accurately stated in the form Every
numerical equation of the n th degree has n numerical roots. "With reference to this proposition, there appears little doubt that the most direct and scientific proof is one founded on the treatment of

imaginary expressions or complex numbers of the kind considered in Chapter XVII.

bers

The

first

idea of the representation of complex

num-

by points in a plane is due to Argand, who in 1806 published anonymously in Paris a work entitled Essai sur une maniere de reprisenter les quantites imaginaires dans les constructions geometriques.

This

writer some years later gave an account of his researches in Gergonne's Annales. Notwithstanding the publicity thus given by Argand to his new methods, they attracted but little notice, and appear to have been
discovered independently several years later by Warren in England and Mourey in France. These ideas were developed by Gauss in his

Notes.

443

works published in 1831 and by Cauchy, who applied them to the With reference to the proof of the important theorem of Art. 194. proposition now under discussion, the proof which we have given at
;

the close of Art. 195

is
1

is to be found in Argand's original memoir, and reproduced by Cauchy with some modifications in his Exercices

d Analyse. A

proof in many respects similar was given by Mourey. Before the discovery of the geometrical treatment of complex numbers several mathematicians occupied themselves with the proof the nature of the roots of equations.
is

blem

An

account of their

researches
riques.

given by Lagrange

in

Note IX.

of his Equations

Numc-

The

inquiries of these investigators,

among whom we may

mention D'Alembert, Descartes, Euler, Foncenex, and Laplace, referred only to equations with rational coefficients; and the object in view was, assuming the existence of factors of the form x - a, x (5,

&c., to

show that the

roots a,

/?,

&c.,

nary quantities of the type a + b<\/ solution of an equation with real numerical
to an imaginary root of

were all either real or imagi- 1 in other that the

;

words,

coefficients cannot give rise

,

in

any form except the known form a + b */- 1 which a and b are real quantities. For the proof of this proposition the method employed in general was to show that, in case of an
its

equation whose degree contained 2 in any power &, the possibility of having a real quadratic factor might be made to depend on the
solution of an equation
;

whose degree contained 2

in the

power k

only and by this process to reduce the problem finally to depend on the known principle that every equation of odd degree with real coefficients has a real root. Lagrange's own investigations on this subject,

given in Note X. of the work above referred to, related, like those of his predecessors, to equations with rational coefficients, and are founded ultimately on the same principle of the existence of a real
root in an equation of odd degree with real coefficients. As resting on the same basis, viz., the existence of a real root in

an equation of odd degree,

may

methods

of considering this

problem

be noticed two recently published one by the late Professor Clif-

ford (see his Mathematical Papers, p. 20, and Cambridge Philosophical 1876), and the other by Professor Malet Society's Proceedings, II.
,

(Transactions of the Royal Irish Academy, vol. xxvi., p. 453, 1878). th Starting with an equation of the 2m degree, botli writers employ of to obtain an equation of the method elimination Sylvester's dialytic

degree m('2m

1)

on whose solution the existence

of a root of the

444

Notes.

proposed equation is shown to depend and since the number m (2m 1) 2 the the contains the factor once less often than number 2w, problem
;

is

reduced ultimately to depend, as in the methods above mentioned, of a root in an equation of odd degree. The two is be effected are between which the elimination to supposed equations

on the existence
of the degrees

and

and the only difference between the two

modes of proof consists in the manner of arriving at these equations. In Professor Malet's method they are found by means of a simple
transformation of the proposed equation, while Professor Clifford obtains them by equating to zero the coefficients of the remainder when

the given polynomial

of these coefficients are given in

The forms divided by a real quadratic factor. Ex. 31, p. 286 and it will be readily observed that the elimination of fi from the equations obtained by
is
;

making rQ and m(2m - 1).

r^

vanish will furnish an equation in a of the degree

INDEX.
Abel, 435.
Bombelli, 434.

Addition of determinants, 248.

Algebraical equations,
2, 197.

Bring, 436.
Brioschi's differential equation, 304. Budan theorem of Fourier and, 17'2.
:

their solution, 103.

solution of cubic, 106. of biquadratic, 119.

Alkbayyami, 433.
Alternants, 283.

Cardan: solution of cubic, 106. his relations with Tartaglia, 434.

Cauchy
:

his theorem, 431.

Approximation to numerical roots Newton's, 207.

Horner's, 209. Lagrange's, 223.
Arabians, 433. Argand, 442.

on determinants, 441.
Cayloy
:

number

solution of cubic, 369. of covariants and

inva-

riants of cubic, 369.

solution of quartic, 376.

results of

Tsehimhausen's transfor-

Argument

of complex variable, 422. variation of, 428.

mation, 389. Circulants, 284.

Clebsch, referred
Colla, 433.
to,

Arrays, square, 232. rectangular, 259.

359, 397, 436.

Cogredient defined, 357.

Commensurable
Ball,

roots

theorem, 198.

on quadratic

factors of sextic co-

Complex

variable, 425.

variant of quartic, 372.

Baltzer, 441.

Continuants, 285.
Continuity, rational integral function, 9. function of complex variable, 427. Covariants definitions, 338.
:

Ben Musa, 433. Bezout's method

411.

of elimination, 324.

Binary forms transformed to ternary,

Binomial
coefficients, 68.
:

formation

of,

339.

properties of, 341.

Binomial equations

solution of, 90. leading general properties of, 92. solution by circular functions, 98.
of,

formation by operator D, 344. Roberts' theorem relating to, 346. homographic transformation applied to, 347. properties derived formation, 351.

Biquadratic, 73. Euler's solution

Ferrari's, 127.

by

linear trans-

119.

Descartes', 131.

to their forpropositions relating mation, 354, 356, 358.

transformed to reciprocal form, 133. solved by symmetric functions, 137

equation of differences, 140. nature of ite roots, 142, 194.

of cubic, 366.
their

number, 369.

of quartic, 371. their number, 379.

446
Cube
roots of unity, 43.

Index.
Equation whose roots are any powers of
roots of given equation, 78. Equations linear, solution of, 262.

Cubic, 71.

equation of differences, 81. criterion of nature of its roots, 84. Cardan's solution
as difference of
of, 106.

linear homogeneous, 263. Euler: solution of biquadratic, 119.

two cubes, 109.

111.

solved

by symmetric functions,

method of elimination, 322. publication of his Algebra, 434.

Ferrari
:

homograpbic relation between two

roots, 118.

solution of biquadratic, 127.

covariants and invariants, 366, 369.

Florido, 433.

transformed

by

Tscbirnhausen's

method, 389.

Fourier: his theorem, 172, 440. the theorem applied to imaginary

roots, 177.

De Gua

rule for finding imaginary roots, 180.

:

corollaries

from the theorem, 180.

Galois, 436.

Derived functions,

8.

Graphic representation

graphic representation of, 152. in terms of the roots, 155.

Descartes
rule of signs, 28, 30, 180. solution of the biquadratic, 131.
:

[13. of polynomial, of derived functions, 152.

:

of imaginary quantities, 422. Greatheed solution of biquadratic, 134.

:

his improvements in Algebra, 434. Determinants definitions, 229.

:

Harley, 436.

Hermite

his theorem relating to limits

propositions relating to, 234-255.

of roots, 403.

minor determinants, 239.

development
addition
of,
of,

on the reduction of the quintic, 436.

Hessian: of cubic, 340, 367.
of quartic, 346, 371. form in general, 357. of quartic expressed by factors of
its

239, 244, 245, 247.

248.

multiplication of, 255.

reciprocal determinants, 264. symmetric determinants, 267.

sextic covariant, 374.

skew and skew -symmetric, 269.

miscellaneous examples in, 276. note on their history, 441.
Dialytic method of elimination, 323. Discriminants, 331.
Divisors,

Homogeneous
quadratic

linear equations, 263.

sum of expressed as squares, 401. Homographic transformation, 75. reduced to double linear transformation, 349.

Newton's method

of, 199.

Homographic

relation

between roots of

Elimination, 318.

a cubic, 118, 399.

by symmetric functions, 319. Euler's method, 322. Sylvester's method, 323.

'

Homer

his

method of solving nume-

rical equations, 209.

contraction of the process, 217.

process applied to cases where roots are nearly equal, 220.
his

Bezout's method, 324. the common method, 329.

Equal

roots, 25.

condition for in cubic, 84. in biquadratic, 142.

improvements in solution of numerical equations, 439.

roots, 21.

Imaginary
:

determination

of, 157.

enter equations in pairs, 26.

Equation of squared differences

of cubic, 81. of the general equation, 84. of biquadratic, 140.

Imaginary quantities

graphic representation, 422. addition and subtraction of, 423.

multiplication and division of, 424.

447
Invariants
:

definition?, :>38, 349.

of,

formation

339.

Numerical equations, 2, 197. commensurable roots of, 19S.

multiple roots of, 157, 204. methods of approximation,
2>7,

properties of, 341. modes of generation, 355, 356.

of cubie, 369.

of quartic, 371, 379. of the form K r

209, 22?. note on the solution

,

\H

377.

of, 437. note on the proposition that every equation has a root, 442.

Jacobian, defined, 359. Jernrnl, 436.

Order of Symmetric Functions, 299.

Jordan, 436.

Polynomials

general properties,
of, 8.

5, 6.

change of form

Lagiauge
223.

method of approximation,

continuity of, 9.

graphic representation

of, 13.

on equation of differences, 140. on solution of equations, 435. his treatise on Numerical Equations
referred
to,

maxima and minima,

Quartic
:

17.

covariants and invariants, 371.

141, 438, 439, 443.

Laplace
244.

development of determinant,

expressed by quadratic factors of sextic covariant, 374.

resolution of, 376. number of its covariants and invariants, 379.

proof that equation has a root, 443. Leonardo, 443. Limits of roots definitions, 163.
:

propositions
limits,

relating to 163, 164, 168. Newton's method, 168.

tive roots, 169.

transformed by Tschirnhausen, 390. two transformable into each other,

396.
solution of special form, 102. Quintic reduction to trinomial form, 394.
:

superior

inferior limits,
.

and limits of nega-

to

sum

of three fifth powers, 395.

Limiting equations, 170. Linear equations solution

:

of, 262.

impossibility of its solution dicals, 435.

by

ra-

homogeneous, 263. Lucas de Burgo, 433.

Quotient and remainder: when polynomial is divided by binomial, 10. when one polynomial is divided by
another, 286. general forms of the coefficients when polynomial of even degree
is

Maxima and minima,

:

17, 153.

Minor determinants, 239. Modulus linear transformation, 349.

of complex numbers, 422. Multiple roots, 156, 157.

divided
:

by a

quadratic, 286.

Reality of roots

of cubic, 84.

determination

of,

1 ">7.

Multiplication of determinants, 255. of complex numbers, 424.

of biquadratic, 142. in general, 193.

Reciprocal determinants, 264. Reciprocal roots and reciprocal equations,

Newton's method of finding limits, 168,

IbO.

62.

method
of

approximation
roots, 207.

of divisors, 199. to numerical

solution of reciprocal equations, 90. transformation of biquadratic to

reciprocal form, 133. Rectangular arrays, 259.

theorem on sums
roots,

of

powers

of

Removal of terms,
of three terms

67.

289.

by Tschirnhausen's

Numbers, complex, 108, 425.

transformation, 393.

448
Resultant of two equations, 318.
properties of, 320.

Index.
Sums
of powers of roots Xewton's theorem on, 289.
:

Roberts

on an equation derived from

t\vo cubics, 116.

in terms of coefficients, 293. coefficients expressed by, 294.

on equation of squared differences

of biquadratic, 142. identical relation, 151.

example on quartic and quintic, on source of co variant, 339. theorem on covariants, 346. on product of covariants, 362. multinomial theorem, 362.
Rolle's theorem, 155.

95.

operation involving, 309. Sylvester: method of elimination, 323. reduction of quintic to sum of three fifth powers, 396.

theorem relating to
pressed as

quadric

ex-

of squares, 402. forms of Sturm's remainders, 406.

sum

Symmetric functions:
theorems relating

definitions, 46.
to, 53.
of, 76.

Roots

theorems relating
of,

to, 19.

transformation by means

imaginary, 21.

number

22.

expressed rationally coefficients, 291.

in

terms of

equal, 25. Descartes' rule for positive, 28. for negative and imaginary, 30.

order and weight of, 53, 299.

calculation of, 300. by Brioschi's equation, 305.

how

related to coefficients, 35.

derived from given one, 306.

applied to elimination, 319. of the roots of two equations, 334.

cube roots of unity, 43.

symmetric functions
multiple, 156, 204.
limits of, 163.

of, 46,

289.

separation of, 172. commensurable, 198.

Tabulation of functions, 12. Tartaglia, 433.

Transformation

two equations, 333. Hermite's theorem on limits of, 403.

to

common

of equations, 60. of cubic, 71. of biquadratic, 73.

:

Routh
Rule
:

Cauchy's theorem concerning, 431. examples in determinants, 288.

:

homographic, 75.
functions, 76. in general, 80. reduction of homographic to double

by symmetric

Descartes', 28. of signs for determinant, 232.

DeG-ua's, 180.

linear, 349. linear, applied to covariants, 351.

Salmon's Modern Higher Algebra referred to, 288, 359, 436, 442. Scipio Ferreo, 433.

theorem relating

to, 385.

Tschirnhausen's, 387. geometrical, 411.

of binary to ternary forms, 411. Tschirnhausen's transformation, 387.

Separation of roots, 172. Serret's Algebra referred

394, 399. Sextic covariant of quartic, 372. Simpson, 434.

to,

applied to cubic, 389. to quartic, 390.

Skew and skew-symmetric

determi-

reduction

of

cubic

to

binomial

nants, 269. Special roots of binomial equations, 95. Sturm: his theorem, 181.
for equal roots, 186. application of theorem, 189.

form, 391.
of quartic to trinomial form, 392. of quintic to trinomial form, 394.

Vandermonde, 435, 441.

Variable, complex, 425.
Vieta, 437.

leading coefficients of functions expressed as determinants, 312. Sylvester's forms of his remainders, 406.

Wantzel, 435.

Weight of symmetric

funoti' in?,

.">.'',

23D,

68

PLEASE DO NOT REMOVE

CARDS OR SLIPS FROM THIS POCKET

UNIVERSITY OF

TORONTO LIBRARY

Physical

&

Applied ScL

gMMgRj&yyf

$JIM$